# Load Balancing

Load balancing is the distribution of network traffic across multiple servers or other application-supporting resources to improve availability, scalability, security, and performance.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Introducing mTLS for Network Load Balancer

DevFeed: [Introducing mTLS for Network Load Balancer](<https://devfeed.tech/articles/introducing-mtls-for-network-load-balancer-144602.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/networking-and-content-delivery/introducing-mtls-for-network-load-balancer/>)

Author: Rizwan Mushtaq

Published: 2026-11-19T22:45:08Z

Content type: release

Language: en

Sources: [Networking & Content Delivery](<https://devfeed.tech/sources/networking-content-delivery.md>)

Topics: [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [TLS (Transport Layer Security)](<https://devfeed.tech/topics/tls.md>), [certificates](<https://devfeed.tech/topics/certificates.md>), [Amazon CloudWatch](<https://devfeed.tech/topics/amazon-cloudwatch.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [advanced-300](<https://devfeed.tech/tags/advanced-300.md>), [allowed](<https://devfeed.tech/tags/allowed.md>), [amazon-cloudwatch](<https://devfeed.tech/tags/amazon-cloudwatch.md>), [announcements](<https://devfeed.tech/tags/announcements.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [aws](<https://devfeed.tech/tags/aws.md>), [certificate](<https://devfeed.tech/tags/certificate.md>), [certificates](<https://devfeed.tech/tags/certificates.md>), [cloudwatch](<https://devfeed.tech/tags/cloudwatch.md>), [elastic-load-balancing](<https://devfeed.tech/tags/elastic-load-balancing.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [load-balancer](<https://devfeed.tech/tags/load-balancer.md>), [logs](<https://devfeed.tech/tags/logs.md>), [metrics](<https://devfeed.tech/tags/metrics.md>), [monitor](<https://devfeed.tech/tags/monitor.md>), [mtls](<https://devfeed.tech/tags/mtls.md>), [mutual-tls](<https://devfeed.tech/tags/mutual-tls.md>), [network](<https://devfeed.tech/tags/network.md>), [spiffe](<https://devfeed.tech/tags/spiffe.md>), [store](<https://devfeed.tech/tags/store.md>), [tls](<https://devfeed.tech/tags/tls.md>), [trust](<https://devfeed.tech/tags/trust.md>)

### AI overview

Network Load Balancer (NLB) can verify client certificates at a TLS listener, so only clients that pass trust, optional identity filters, and revocation checks reach backend targets; targets receive no client certificate data. Configure verify mode with a trust store built from a CA bundle in Amazon S3, optionally add a CRL, and optionally filter on certificate CN, OU, DNS SAN, or URI SAN (including SPIFFE IDs). Filters use exact or prefix matching, are OR-combined, and apply per listener. NLB checks trust chains, certificate expiry and purpose, then filters and revocation; expired certificates can be allowed through an advanced setting. CRLs are imported when added and must be added again when updated. Access logs record authentication failures, while CloudWatch metrics cover handshakes, errors, latency, and certificate expiration. The article also describes setup prerequisites, testing, trust store sharing, limits, and configuration through console, CLI, SDKs, or CloudFormation.

### Source excerpt

Network Load Balancer now offloads mutual TLS for you. Learn how to attach a trust store, validate client certificates against your revocation lists, restrict access by CN, OU, or SPIFFE ID with client certificate filtering, and monitor connections with access logs and Amazon CloudWatch metrics.

## F5 BIG-IP Under Pressure: Six Critical Vulnerabilities Expose the Traffic Management

DevFeed: [F5 BIG-IP Under Pressure: Six Critical Vulnerabilities Expose the Traffic Management](<https://devfeed.tech/articles/f5-big-ip-under-pressure-six-critical-vulnerabilities-expose-the-traffic-management-143607.md>)

Original publisher: [Read original article](<https://thecyberthrone.in/2026/11/19/f5-big-ip-under-pressure-six-critical-vulnerabilities-expose-the-traffic-management/>)

Author: PravinKarthik

Published: 2026-11-19T11:22:59Z

Content type: article

Language: en

Sources: [Cyber Newsroom](<https://devfeed.tech/sources/cyber-newsroom.md>)

Topics: [cve remediation](<https://devfeed.tech/topics/cve-remediation.md>), [Exposure Management](<https://devfeed.tech/topics/exposure-management.md>), [software security audit](<https://devfeed.tech/topics/software-security-audit.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>)

Tags: [application-security](<https://devfeed.tech/tags/application-security.md>), [asset-inventory](<https://devfeed.tech/tags/asset-inventory.md>), [big-ip](<https://devfeed.tech/tags/big-ip.md>), [big-ip-apm](<https://devfeed.tech/tags/big-ip-apm.md>), [buffer-overflow](<https://devfeed.tech/tags/buffer-overflow.md>), [business-applications](<https://devfeed.tech/tags/business-applications.md>), [code-execution](<https://devfeed.tech/tags/code-execution.md>), [critical](<https://devfeed.tech/tags/critical.md>), [critical-infrastructure](<https://devfeed.tech/tags/critical-infrastructure.md>), [critical-vulnerabilities](<https://devfeed.tech/tags/critical-vulnerabilities.md>), [cve-2026-116953](<https://devfeed.tech/tags/cve-2026-116953.md>), [cve-2026-120285](<https://devfeed.tech/tags/cve-2026-120285.md>), [cve-2026-120292](<https://devfeed.tech/tags/cve-2026-120292.md>), [cve-2026-120722](<https://devfeed.tech/tags/cve-2026-120722.md>), [cve-2026-121805](<https://devfeed.tech/tags/cve-2026-121805.md>), [cve-2026-123637](<https://devfeed.tech/tags/cve-2026-123637.md>), [f5](<https://devfeed.tech/tags/f5.md>), [security](<https://devfeed.tech/tags/security.md>), [traffic](<https://devfeed.tech/tags/traffic.md>), [traffic-management](<https://devfeed.tech/tags/traffic-management.md>), [vulnerabilities](<https://devfeed.tech/tags/vulnerabilities.md>)

### AI overview

F5 disclosed six critical BIG-IP stack-based buffer overflow vulnerabilities, each scored 9.8 under CVSS v3.1 and 9.3 under v4.0. Exposure depends on specific iRule commands or, for CVE-2026-123637, an APM per-request policy combining Client Information and Endpoint Security agents; the article says affected functionality may be triggered by specially crafted traffic from an unauthenticated remote attacker. TMM termination can disrupt traffic processing, with potential remote code execution consequences. Teams are advised to inspect iRules and APM policies, map affected assets, modules and versions against F5's individual advisories, then apply relevant engineering hotfixes or fixed releases. The article cautions that engineering hotfixes receive less extensive quality assurance than scheduled releases, so teams should validate compatibility and operational impact. It also recommends reviewing TMM crashes, logs, traffic interruptions and failover activity, and assessing exposure through configuration and business dependencies rather than software version alone.

### Source excerpt

F5 BIG-IP Under Pressure: Six Critical Vulnerabilities Expose the Traffic Management Enterprise infrastructure has a dependency that often receives less attention than the applications it protects. F5 BIG-IP is one such component. It sits in the path of application traffic, supporting ... Read more Published Date: Nov 19, 2026 (2 hours, 42 minutes ago) Vulnerabilities has been mentioned in this article. CVE-2026-123637 CVE-2026-123633 CVE-2026-123007 CVE-2026-121805 CVE-2026-120722 CVE-2026-120292 CVE-2026-120285 CVE-2026-116953 CVE-2026-114938

## Who's Delivering GTA 6? A RIPE Atlas Measurement Story

DevFeed: [Who's Delivering GTA 6? A RIPE Atlas Measurement Story](<https://devfeed.tech/articles/who-s-delivering-gta-6-a-ripe-atlas-measurement-story-143194.md>)

Original publisher: [Read original article](<https://labs.ripe.net/author/emileaben/whos-delivering-gta-6-a-ripe-atlas-measurement-story/>)

Author: Emile Aben

Published: 2026-11-18T13:48:32Z

Content type: article

Language: en

Sources: [RIPE Labs](<https://devfeed.tech/sources/ripe-labs.md>)

Topics: [Domain Name System (DNS)](<https://devfeed.tech/topics/dns-resolution-process.md>), [Internet](<https://devfeed.tech/topics/internet.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Internet Traffic](<https://devfeed.tech/topics/internet-traffic.md>)

Tags: [akamai](<https://devfeed.tech/tags/akamai.md>), [alibaba-cloud](<https://devfeed.tech/tags/alibaba-cloud.md>), [amazon-cloudfront](<https://devfeed.tech/tags/amazon-cloudfront.md>), [atlas](<https://devfeed.tech/tags/atlas.md>), [balancing](<https://devfeed.tech/tags/balancing.md>), [cname](<https://devfeed.tech/tags/cname.md>), [content-delivery-networks](<https://devfeed.tech/tags/content-delivery-networks.md>), [dns](<https://devfeed.tech/tags/dns.md>), [gta-6](<https://devfeed.tech/tags/gta-6.md>), [measurement](<https://devfeed.tech/tags/measurement.md>), [measurements](<https://devfeed.tech/tags/measurements.md>), [playstation](<https://devfeed.tech/tags/playstation.md>), [research](<https://devfeed.tech/tags/research.md>), [ripe](<https://devfeed.tech/tags/ripe.md>), [xbox](<https://devfeed.tech/tags/xbox.md>)

### AI overview

RIPE Atlas measurements of GTA 6 pre-load traffic show PlayStation and Xbox using DNS steering and multiple CDNs to distribute downloads. The same PlayStation hostname resolved through different chains to Fastly in Los Angeles, Google Cloud in Tokyo, CloudFront in Istanbul, and Alibaba Cloud in Shanghai, illustrating location-based routing. Across the measured period, the observed server IP pools stayed stable. Both consoles used Akamai and Fastly; measurements also found Google and Amazon for PlayStation and G-Core for Xbox. About 10-30% of measured paths traversed an IXP, with median latency of 8 ms on those paths versus 10 ms on others; the article cautions this does not show IXPs cause lower latency. RIPE Atlas coverage is Europe-heavy and may bias CDN shares, and the measurements do not reveal network traffic loads or explain reported congestion.

### Source excerpt

How do you deliver a 200 GB game to millions of players? As GTA 6 pre-loads got underway last week, we turned to RIPE Atlas to examine how DNS and content delivery networks help PlayStation and Xbox share the load.

## CheckCo: Checkpoint-Coordinated Load Balancingfor Skewed Event Streams at Scale

DevFeed: [CheckCo: Checkpoint-Coordinated Load Balancingfor Skewed Event Streams at Scale](<https://devfeed.tech/articles/checkco-checkpoint-coordinated-load-balancingfor-skewed-event-streams-at-scale-140388.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22476>)

Author: Muhammad Anis Uddin Nasir, Reza Sadraei

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Event-Streaming](<https://devfeed.tech/topics/event-streaming.md>), [streaming data analytics](<https://devfeed.tech/topics/streaming-data-analytics.md>)

Tags: [checkpoint](<https://devfeed.tech/tags/checkpoint.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [event-streams](<https://devfeed.tech/tags/event-streams.md>), [load-balancing](<https://devfeed.tech/tags/load-balancing.md>), [partitioning](<https://devfeed.tech/tags/partitioning.md>), [routing](<https://devfeed.tech/tags/routing.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>), [workers](<https://devfeed.tech/tags/workers.md>)

### AI overview

CheckCo uses checkpoint barriers to adapt routing to skewed event streams without prior key-frequency statistics. Workers count events per key locally; at each barrier, counts are merged to split keys roughly in proportion to their estimated frequency and pack those splits onto workers for the next checkpoint window. The authors derive bounds on load imbalance and total key fan-out. On a production pipeline handling over 100 billion events daily, the most-loaded worker received 9.8% more events than average over a day (imbalance ratio 1.10), compared with 8.03 for key-based routing in simulation. In simulation, CheckCo used less than half the fan-out of earlier partitioners with comparable balance. Routing reflects distribution shifts from the next checkpoint, and the authors report balanced load through job restarts.

### Source excerpt

arXiv:2611.22476v1 Announce Type: new Abstract: Event streams are rarely uniform: a few dominant keys can generate orders of magnitude more events than the rest, overloading the workers that receive them and capping throughput. We present CheckCo, a checkpoint-coordinated load-balancing algorithm built on a dynamic oracle that estimates per-key frequencies from the live stream, with no prior statistics. Workers count events per key locally; at each checkpoint barrier, the counts are merged into a partitioning function that divides each key into a number of splits roughly proportional to its estimated frequency and packs the splits onto workers, and this function routes the next checkpoint window. We derive approximation bounds on the resulting load imbalance from classical scheduling results and bound its total key fan-out. On a production pipeline processing over 100 billion events per day, the most loaded worker receives only 9.8% more events than the average over a day (imbalance ratio 1.10), against 8.03 for key-based routing in simulation; in simulation, CheckCo also needs less than half the fan-out of earlier partitioners that reach comparable balance. Shifts in the key distribution are reflected in routing from the next checkpoint on, and the load stays balanced through job restarts.

## Scaling Azure Application Gateway for traffic spikes

DevFeed: [Scaling Azure Application Gateway for traffic spikes](<https://devfeed.tech/articles/scaling-azure-application-gateway-for-traffic-spikes-144741.md>)

Original publisher: [Read original article](<https://www.gremlin.com/blog/scaling-azure-application-gateway-for-traffic-spikes>)

Author: Andre Newman

Published: 2026-11-18T00:00:00Z

Content type: article

Language: en

Sources: [Gremlin Blog](<https://devfeed.tech/sources/gremlin-blog.md>)

Topics: [Azure](<https://devfeed.tech/topics/azure.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [gateway](<https://devfeed.tech/topics/gateway.md>)

Tags: [autoscaling](<https://devfeed.tech/tags/autoscaling.md>), [azure](<https://devfeed.tech/tags/azure.md>), [black-friday](<https://devfeed.tech/tags/black-friday.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [gateway](<https://devfeed.tech/tags/gateway.md>), [scaling](<https://devfeed.tech/tags/scaling.md>), [spikes](<https://devfeed.tech/tags/spikes.md>), [use-cases](<https://devfeed.tech/tags/use-cases.md>)

### AI overview

Azure Application Gateway v2 can use manual scaling (a fixed 1-125 instances) or autoscaling (a minimum of 0-100 and an optional maximum up to 125, default 10). Fixed capacity can constrain traffic during unexpected surges and leave unused capacity running when traffic falls. Azure scales using capacity units based on the highest of compute, persistent connections, or throughput; each instance handles up to 10 units. Because new instances take three to five minutes to provision, the article suggests starting with a minimum of at least two and sizing from recent peak usage. It recommends a maximum of 125 for high-traffic gateways, a /24 subnet for v2, and alerts at 75% of recent peak capacity; workloads needing more than 125 instances can route across gateways with Traffic Manager or Front Door. Autoscaling can be enabled without redeploying through the portal, CLI, Bicep, or Terraform. Azure Resource Graph can find fixed-capacity gateways across accessible subscriptions; Gremlin also flags gateways lacking autoscaling and monitors for configuration regressions.

### Source excerpt

Fixed-capacity Azure Application Gateways limit your total traffic capacity. Learn how autoscaling works, how to enable it, and how to find gateways without it.

## Why the Holidays are Gaming's Biggest Season and Its Riskiest

DevFeed: [Why the Holidays are Gaming's Biggest Season and Its Riskiest](<https://devfeed.tech/articles/why-the-holidays-are-gaming-s-biggest-season-and-its-riskiest-143979.md>)

Original publisher: [Read original article](<https://www.fastly.com/blog/why-holidays-are-gamings-biggest-season-and-riskiest/>)

Author: Ashley Hurwitz

Published: 2026-11-18T00:00:00Z

Content type: article

Language: en

Sources: [Fastly Blog](<https://devfeed.tech/sources/fastly-blog.md>)

Topics: [observability in gaming](<https://devfeed.tech/topics/observability-in-gaming.md>), [observability](<https://devfeed.tech/topics/observability.md>), [rate-limiting](<https://devfeed.tech/topics/rate-limiting.md>), [Internet Traffic](<https://devfeed.tech/topics/internet-traffic.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>)

Tags: [account-takeover](<https://devfeed.tech/tags/account-takeover.md>), [apis](<https://devfeed.tech/tags/apis.md>), [attack](<https://devfeed.tech/tags/attack.md>), [attacks](<https://devfeed.tech/tags/attacks.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [bot-detection](<https://devfeed.tech/tags/bot-detection.md>), [bot-protection](<https://devfeed.tech/tags/bot-protection.md>), [bots](<https://devfeed.tech/tags/bots.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [credential-stuffing](<https://devfeed.tech/tags/credential-stuffing.md>), [ddos](<https://devfeed.tech/tags/ddos.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-cloud](<https://devfeed.tech/tags/edge-cloud.md>), [fastly](<https://devfeed.tech/tags/fastly.md>), [fraud](<https://devfeed.tech/tags/fraud.md>), [fraud-prevention](<https://devfeed.tech/tags/fraud-prevention.md>), [gaming](<https://devfeed.tech/tags/gaming.md>), [holidays](<https://devfeed.tech/tags/holidays.md>), [industry-insights](<https://devfeed.tech/tags/industry-insights.md>), [traffic](<https://devfeed.tech/tags/traffic.md>)

### AI overview

Holiday gaming traffic surges can conceal credential stuffing, account takeover, gift card fraud, or DDoS activity, while capacity failures and attacks may look alike. The excerpt says attackers target gaming APIs as well as web portals, and that peak volumes make threats harder to distinguish. It recommends absorbing surges with edge load balancing, correlating real-time logs with availability and login signals, rate limiting requests before they reach origin, and using automated bot detection and edge WAF rules. These controls need to distinguish players from bots: overly rigid rules can block legitimate logins or registrations. The described goal is to keep the platform available and catch attacks without slowing or locking out players.

### Source excerpt

Holiday traffic surges in gaming hide the attacks riding along with them. Learn how to protect capacity and catch credential stuffing without slowing down players.

## Distribute Amazon RDS traffic between read replicas using Network Load Balancer

DevFeed: [Distribute Amazon RDS traffic between read replicas using Network Load Balancer](<https://devfeed.tech/articles/distribute-amazon-rds-traffic-between-read-replicas-using-network-load-balancer-138652.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/database/distribute-amazon-rds-traffic-between-read-replicas-using-network-load-balancer/>)

Author: Vijay Chakraborty

Published: 2026-11-17T15:51:53Z

Content type: tutorial

Language: en

Sources: [AWS Database Blog](<https://devfeed.tech/sources/aws-database-blog.md>)

Topics: [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Amazon RDS](<https://devfeed.tech/topics/amazon-rds.md>), [read-replica](<https://devfeed.tech/topics/read-replica.md>), [Virtual Private Network](<https://devfeed.tech/topics/vpn.md>), [Cloud Networking](<https://devfeed.tech/topics/cloud-networking.md>), [Latency](<https://devfeed.tech/topics/latency.md>), [connection draining](<https://devfeed.tech/topics/connection-draining.md>)

Tags: [advanced-300](<https://devfeed.tech/tags/advanced-300.md>), [affinity](<https://devfeed.tech/tags/affinity.md>), [amazon-aurora](<https://devfeed.tech/tags/amazon-aurora.md>), [amazon-cloudwatch](<https://devfeed.tech/tags/amazon-cloudwatch.md>), [amazon-ec2](<https://devfeed.tech/tags/amazon-ec2.md>), [amazon-elastic-container-service](<https://devfeed.tech/tags/amazon-elastic-container-service.md>), [amazon-elastic-kubernetes-service-amazon-eks](<https://devfeed.tech/tags/amazon-elastic-kubernetes-service-amazon-eks.md>), [amazon-eventbridge](<https://devfeed.tech/tags/amazon-eventbridge.md>), [amazon-rds](<https://devfeed.tech/tags/amazon-rds.md>), [amazon-route-53](<https://devfeed.tech/tags/amazon-route-53.md>), [amazon-s3](<https://devfeed.tech/tags/amazon-s3.md>), [aws-cloudformation](<https://devfeed.tech/tags/aws-cloudformation.md>), [aws-direct-connect](<https://devfeed.tech/tags/aws-direct-connect.md>), [aws-lambda](<https://devfeed.tech/tags/aws-lambda.md>), [aws-privatelink](<https://devfeed.tech/tags/aws-privatelink.md>), [aws-site-to-site-vpn](<https://devfeed.tech/tags/aws-site-to-site-vpn.md>), [elastic-load-balancing](<https://devfeed.tech/tags/elastic-load-balancing.md>), [health](<https://devfeed.tech/tags/health.md>), [lambda](<https://devfeed.tech/tags/lambda.md>), [load-balancer](<https://devfeed.tech/tags/load-balancer.md>), [read-replicas](<https://devfeed.tech/tags/read-replicas.md>), [technical-how-to](<https://devfeed.tech/tags/technical-how-to.md>), [traffic](<https://devfeed.tech/tags/traffic.md>)

### AI overview

An NLB can provide one stable endpoint for distributing connections across Amazon RDS read replicas, with health checks and Availability Zone affinity. An EventBridge rule invokes a Lambda function every minute to resolve replica DNS names and synchronize their IPs into the NLB target group; S3 JSON state and retry logic help avoid premature deregistration during transient DNS failures. With zonal DNS affinity enabled and cross-zone load balancing disabled, traffic can remain AZ-local in steady state, reducing inter-AZ transfer. The endpoint also supports hybrid access over Direct Connect or Site-to-Site VPN without PrivateLink or Route 53 Resolver endpoints when a custom domain is unnecessary. The approach works with RDS engines that support read replicas, but the NLB balances connections rather than queries, so long-lived connections can limit redistribution. In EKS, CoreDNS placement can undermine zonal affinity; NodeLocal DNSCache is suggested. Deregistration delay allows existing connections to drain.

### Source excerpt

Learn how to put Amazon RDS read replicas behind a Network Load Balancer with health-aware, Availability Zone affinity routing. An AWS Lambda function keeps the load balancer target group in sync with replica IP addresses, giving clients a single stable endpoint over Direct Connect or Site-to-Site VPN without AWS PrivateLink overhead.

## Virtual Private Clouds in VMware vSphere Kubernetes Service: The Recommended Network Stack

DevFeed: [Virtual Private Clouds in VMware vSphere Kubernetes Service: The Recommended Network Stack](<https://devfeed.tech/articles/virtual-private-clouds-in-vmware-vsphere-kubernetes-service-the-recommended-network-stack-138605.md>)

Original publisher: [Read original article](<https://blogs.vmware.com/cloud-foundation/2026/11/17/virtual-private-clouds-in-vmware-vsphere-kubernetes-service-the-recommended-network-stack/>)

Author: vmwareblogs

Published: 2026-11-17T14:00:00Z

Content type: article

Language: en

Sources: [VMware](<https://devfeed.tech/sources/vmware-blogs.md>)

Topics: [vcf networking](<https://devfeed.tech/topics/vcf-networking.md>), [Cloud Networking](<https://devfeed.tech/topics/cloud-networking.md>), [VMware Cloud Foundation](<https://devfeed.tech/topics/vcf-9-1.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [NAT and SNAT](<https://devfeed.tech/topics/nat-and-snat.md>), [container-networking](<https://devfeed.tech/topics/container-networking.md>), [private cloud](<https://devfeed.tech/topics/private-cloud.md>), [vcf automation](<https://devfeed.tech/topics/vcf-automation.md>), [self-service](<https://devfeed.tech/topics/self-service.md>)

Tags: [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-infrastructure](<https://devfeed.tech/tags/cloud-infrastructure.md>), [home-page](<https://devfeed.tech/tags/home-page.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kubernetes-nodes](<https://devfeed.tech/tags/kubernetes-nodes.md>), [kubevm](<https://devfeed.tech/tags/kubevm.md>), [network](<https://devfeed.tech/tags/network.md>), [network-stack](<https://devfeed.tech/tags/network-stack.md>), [networking](<https://devfeed.tech/tags/networking.md>), [nsx](<https://devfeed.tech/tags/nsx.md>), [private](<https://devfeed.tech/tags/private.md>), [private-cloud](<https://devfeed.tech/tags/private-cloud.md>), [recommended](<https://devfeed.tech/tags/recommended.md>), [service](<https://devfeed.tech/tags/service.md>), [supervisor](<https://devfeed.tech/tags/supervisor.md>), [switch](<https://devfeed.tech/tags/switch.md>), [vcf](<https://devfeed.tech/tags/vcf.md>), [vcf-9-1](<https://devfeed.tech/tags/vcf-9-1.md>), [vcf-networking](<https://devfeed.tech/tags/vcf-networking.md>), [vks](<https://devfeed.tech/tags/vks.md>), [vks-vsphere-kubernetes-service](<https://devfeed.tech/tags/vks-vsphere-kubernetes-service.md>), [vmware-cloud-foundation](<https://devfeed.tech/tags/vmware-cloud-foundation.md>), [vmware-vsphere-kubernetes-service](<https://devfeed.tech/tags/vmware-vsphere-kubernetes-service.md>), [vpc](<https://devfeed.tech/tags/vpc.md>), [vsphere](<https://devfeed.tech/tags/vsphere.md>)

### AI overview

VCF Networking with VPC is the recommended VCF 9.1 stack for production VKS deployments and the required stack for VCF Automation environments. It places cluster nodes on private subnets and exposes needed endpoints through load balancing, with outbound NAT; its Virtual Network Appliance cluster scales to ten active/active nodes. The vSphere Distributed Switch option connects nodes to an existing VLAN and routes VIPs through one active/standby Foundation Load Balancer pair. VPC setup can start with one VLAN and grow to centralized Edge connectivity or Cisco and Arista EVPN/VXLAN integration. It also provides vCenter-based networking and Traceflow, multi-tenancy and self-service automation, and integrations including Infoblox and vDefend. A VLAN-backed VPC option in VCF 9.1.1 preserves the VPC workflow but removes private-network creation. The article says the Distributed Switch remains suited to VMware vSphere Foundation and smaller or lab environments.

### Source excerpt

Enabling the vSphere Supervisor in VMware Cloud Foundation (VCF) involves choosing a network stack for workloads running on vSphere Kubernetes Service (VKS), KubeVM Service, or Container Service - VCF Networking with Virtual Private Cloud (VPC), or the vSphere Distributed Switch. VCF Networking with VPC is the recommended, default option. It keeps Kubernetes nodes off the ... Continued The post Virtual Private Clouds in VMware vSphere Kubernetes Service: The Recommended Network Stack appeared first on VMware Blogs.

## Microsoft outage: A cascading failure that started with an undetected bug

DevFeed: [Microsoft outage: A cascading failure that started with an undetected bug](<https://devfeed.tech/articles/microsoft-outage-a-cascading-failure-that-started-with-an-undetected-bug-136600.md>)

Original publisher: [Read original article](<https://www.networkworld.com/article/4249392/microsoft-outage-a-cascading-failure-that-started-with-an-undetected-bug-2.html>)

Author: Evan Schuman

Published: 2026-11-17T00:02:57Z

Content type: news

Language: en

Sources: [Network World](<https://devfeed.tech/sources/network-world.md>)

Topics: [Post Mortem](<https://devfeed.tech/topics/post-mortem.md>), [reliability-engineering](<https://devfeed.tech/topics/reliability-engineering.md>), [microsoft 365](<https://devfeed.tech/topics/microsoft-365.md>), [test](<https://devfeed.tech/topics/test.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Microsoft](<https://devfeed.tech/topics/microsoft.md>), [Microsoft Teams](<https://devfeed.tech/topics/microsoft-teams.md>)

Tags: [availability](<https://devfeed.tech/tags/availability.md>), [bug](<https://devfeed.tech/tags/bug.md>), [defect](<https://devfeed.tech/tags/defect.md>), [edition-us](<https://devfeed.tech/tags/edition-us.md>), [exchange](<https://devfeed.tech/tags/exchange.md>), [exchange-online](<https://devfeed.tech/tags/exchange-online.md>), [failure](<https://devfeed.tech/tags/failure.md>), [gateway](<https://devfeed.tech/tags/gateway.md>), [global](<https://devfeed.tech/tags/global.md>), [global-outage](<https://devfeed.tech/tags/global-outage.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [high-availability](<https://devfeed.tech/tags/high-availability.md>), [incident](<https://devfeed.tech/tags/incident.md>), [incident-report](<https://devfeed.tech/tags/incident-report.md>), [independent](<https://devfeed.tech/tags/independent.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [isolation](<https://devfeed.tech/tags/isolation.md>), [issue](<https://devfeed.tech/tags/issue.md>), [language-en](<https://devfeed.tech/tags/language-en.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-outage](<https://devfeed.tech/tags/microsoft-outage.md>), [networking](<https://devfeed.tech/tags/networking.md>), [outage](<https://devfeed.tech/tags/outage.md>), [storytype-news](<https://devfeed.tech/tags/storytype-news.md>)

### AI overview

A Microsoft 365 gateway software defect, deployed globally for months without impact, surfaced when Teams failed over to the primary gateway, increasing and changing traffic while nearly 14% of gateway capacity was already unavailable. Microsoft's load tests missed the defect because they used different traffic and SKU mixes and tended to use weaker, lower-core hardware; the report says more cores and newer hardware worsened contention. Requests accumulated, and automatic rerouting spread load across fewer healthy servers, causing a fleet-wide cascade that left more than 80% unresponsive at the low point. Recovery automation and normal rollback, flighting and feature flags could not quickly restore service. The preliminary report did not explain the unrelated capacity issue, and the article notes that Microsoft had yet to provide customers concrete reassurance against recurrence.

### Source excerpt

Following a roughly day-long global outage of many Microsoft services last week, Microsoft on Monday issued a preliminary post-incident report that identified multiple causes for the issues, including the admission that the environment used to test the underlying problematic software update did not match the actual infrastructure it would run on. The report, a summary of which was posted on social media by Microsoft EVP Charles Lamanna, identified some of the factors that cascaded to cause the outage. "A latent defect caused servers to become overwhelmed and unresponsive under certain load conditions," the report said, adding, "this defect was the result of a fix that was prepared specifically for Microsoft 365 networking gateway servers." Its deployment began in July, and, the report said, "for more than three months, the change was deployed worldwide without impact." Microsoft admitted that, although it did test the change before deployment, its test environment didn't match reality. "Our normal load testing did not detect the latent issue because the different traffic and SKU mix used in our test labs," it said. "The contention issue is worsened on machines with more cores and newer hardware; whereas our load testing tended to focus on our weaker hardware SKUs with fewer cores." A surge in traffic caused by unrelated issues also helped trigger the problem and delay it from being fixed, the company explained. "Within Microsoft 365, we maintain two independent and heterogenous network gateway systems designed for isolation and high availability," Microsoft said. "This allows many Microsoft 365 services to fail to/from them during outages or disruptions. Approximately fifteen hours before the outage, Teams failed over from the alternative networking gateway to the primary gateway that also supports Exchange Online. This increased the load on the networking gateway and changed its traffic patterns." To make things worse, the report said, "nearly 14% of server capacit

## Microsoft outage: A cascading failure that started with an undetected bug

DevFeed: [Microsoft outage: A cascading failure that started with an undetected bug](<https://devfeed.tech/articles/microsoft-outage-a-cascading-failure-that-started-with-an-undetected-bug-136867.md>)

Original publisher: [Read original article](<https://www.computerworld.com/article/4249364/microsoft-outage-a-cascading-failure-that-started-with-an-undetected-bug.html>)

Author: Evan Schuman

Published: 2026-11-16T23:54:42Z

Content type: news

Language: en

Sources: [Computerworld](<https://devfeed.tech/sources/computerworld.md>)

Topics: [cloud service outage management](<https://devfeed.tech/topics/cloud-service-outage-management.md>), [reliability-engineering](<https://devfeed.tech/topics/reliability-engineering.md>), [Post Mortem](<https://devfeed.tech/topics/post-mortem.md>), [microsoft 365](<https://devfeed.tech/topics/microsoft-365.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [test](<https://devfeed.tech/topics/test.md>), [Microsoft Teams](<https://devfeed.tech/topics/microsoft-teams.md>)

Tags: [bug](<https://devfeed.tech/tags/bug.md>), [defect](<https://devfeed.tech/tags/defect.md>), [edition-us](<https://devfeed.tech/tags/edition-us.md>), [exchange](<https://devfeed.tech/tags/exchange.md>), [exchange-online](<https://devfeed.tech/tags/exchange-online.md>), [failure](<https://devfeed.tech/tags/failure.md>), [gateway](<https://devfeed.tech/tags/gateway.md>), [global-outage](<https://devfeed.tech/tags/global-outage.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [high-availability](<https://devfeed.tech/tags/high-availability.md>), [incident](<https://devfeed.tech/tags/incident.md>), [incident-report](<https://devfeed.tech/tags/incident-report.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [isolation](<https://devfeed.tech/tags/isolation.md>), [language-en](<https://devfeed.tech/tags/language-en.md>), [load](<https://devfeed.tech/tags/load.md>), [load-testing](<https://devfeed.tech/tags/load-testing.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-365](<https://devfeed.tech/tags/microsoft-365.md>), [microsoft-365-office-suites-productivity-software-vendors-and-providers](<https://devfeed.tech/tags/microsoft-365-office-suites-productivity-software-vendors-and-providers.md>), [microsoft-outage](<https://devfeed.tech/tags/microsoft-outage.md>), [network](<https://devfeed.tech/tags/network.md>), [office-suites](<https://devfeed.tech/tags/office-suites.md>), [outage](<https://devfeed.tech/tags/outage.md>), [productivity-software](<https://devfeed.tech/tags/productivity-software.md>), [storytype-news](<https://devfeed.tech/tags/storytype-news.md>), [vendors-and-providers](<https://devfeed.tech/tags/vendors-and-providers.md>)

### AI overview

Microsoft's preliminary report attributes the outage to a latent bug in a networking gateway update, exposed when Teams traffic shifted to the primary gateway and unrelated capacity loss left it under greater load. Testing missed the defect because lab traffic and hardware differed from production; the problem was worse on newer, higher-core servers. As overloaded servers stopped responding, traffic shifted onto fewer healthy servers, spreading the failure. More than 80% of servers became unresponsive at the low point, and restarting them could trigger the bug again. Automated recovery, rollback, flighting and feature flags could not quickly restore service. Consultants cited the production-testing mismatch and delayed alerting as concerns; Microsoft did not explain the unrelated capacity issue in the excerpt.

### Source excerpt

Following a roughly day-long global outage of many Microsoft services last week, Microsoft on Monday issued a preliminary post-incident report that identified multiple causes for the issues, including the admission that the environment used to test the underlying problematic software update did not match the actual infrastructure it would run on. The report, a summary of which was posted on social media by Microsoft EVP Charles Lamanna, identified some of the factors that cascaded to cause the outage. "A latent defect caused servers to become overwhelmed and unresponsive under certain load conditions," the report said, adding, "this defect was the result of a fix that was prepared specifically for Microsoft 365 networking gateway servers." Its deployment began in July, and, the report said, "for more than three months, the change was deployed worldwide without impact." Microsoft admitted that, although it did test the change before deployment, its test environment didn't match reality. "Our normal load testing did not detect the latent issue because the different traffic and SKU mix used in our test labs," it said. "The contention issue is worsened on machines with more cores and newer hardware; whereas our load testing tended to focus on our weaker hardware SKUs with fewer cores." A surge in traffic caused by unrelated issues also helped trigger the problem and delay it from being fixed, the company explained. "Within Microsoft 365, we maintain two independent and heterogenous network gateway systems designed for isolation and high availability," Microsoft said. "This allows many Microsoft 365 services to fail to/from them during outages or disruptions. Approximately fifteen hours before the outage, Teams failed over from the alternative networking gateway to the primary gateway that also supports Exchange Online. This increased the load on the networking gateway and changed its traffic patterns." To make things worse, the report said, "nearly 14% of server capacit

## Scheduling the removal of the kube-proxy IPVS backend

DevFeed: [Scheduling the removal of the kube-proxy IPVS backend](<https://devfeed.tech/articles/scheduling-the-removal-of-the-kube-proxy-ipvs-backend-136734.md>)

Original publisher: [Read original article](<https://kubernetes.io/blog/2026/11/16/ipvs-removal/>)

Author: Dan Winship

Published: 2026-11-16T18:35:00Z

Content type: news

Language: en

Sources: [Kubernetes Blog](<https://devfeed.tech/sources/kubernetes-blog.md>)

Topics: [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [deprecated](<https://devfeed.tech/topics/deprecated.md>), [control-plane](<https://devfeed.tech/topics/control-plane.md>), [Linux](<https://devfeed.tech/topics/linux.md>), [scalable-architecture](<https://devfeed.tech/topics/scalable-architecture.md>)

Tags: [api](<https://devfeed.tech/tags/api.md>), [backend](<https://devfeed.tech/tags/backend.md>), [backends](<https://devfeed.tech/tags/backends.md>), [control-plane](<https://devfeed.tech/tags/control-plane.md>), [data](<https://devfeed.tech/tags/data.md>), [deprecate](<https://devfeed.tech/tags/deprecate.md>), [deprecation](<https://devfeed.tech/tags/deprecation.md>), [edge](<https://devfeed.tech/tags/edge.md>), [experiment](<https://devfeed.tech/tags/experiment.md>), [feature](<https://devfeed.tech/tags/feature.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kubernetes-services](<https://devfeed.tech/tags/kubernetes-services.md>), [migrate](<https://devfeed.tech/tags/migrate.md>), [network](<https://devfeed.tech/tags/network.md>), [nftables](<https://devfeed.tech/tags/nftables.md>), [performance](<https://devfeed.tech/tags/performance.md>), [plan](<https://devfeed.tech/tags/plan.md>), [planning](<https://devfeed.tech/tags/planning.md>), [proxy](<https://devfeed.tech/tags/proxy.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>)

### AI overview

Kubernetes SIG Network plans to remove kube-proxy's IPVS backend in 1.40 after deprecating it in 1.35. IPVS improved on iptables' service lookup and update costs, but its limited API, lack of batch updates, incomplete Kubernetes Service compatibility, and reliance on iptables for some features made it complex to maintain and prone to scaling limits. For Kubernetes 1.36 or later on Linux kernel 5.13 or later, the article recommends switching to the GA nftables backend, which is feature complete and has performance comparable to or slightly better than IPVS. Change kube-proxy's mode and restart it; adjust any backend-specific options too. Where nftables is unavailable, iptables is an alternative: its rule generation and resync behavior have improved, though the article says it remains slower than IPVS and recommends rebooting nodes after switching.

### Source excerpt

The ipvs backend for kube-proxy was an experiment in providing a service proxy implementation with better performance than the iptables backend. While it succeeded in that goal, the kernel IPVS API has turned out to be a bad match for the Kubernetes Services API, and the ipvs backend was never able to implement all of the edge cases of Kubernetes Service functionality correctly. In Kubernetes 1.29, the project introduced the nftables backend for kube-proxy, which became GA in 1.33. Part of the plan behind adding the nftables backend was the idea that it would eventually replace both the iptables and ipvs backends. Because the ipvs backend in particular is complex, not 100% feature complete, and architecturally quite different from the other two backends (and thus more work to maintain), Kubernetes SIG Network decided to officially deprecate it as of Kubernetes 1.35. Although it currently remains available in kube-proxy (with a warning about its deprecation), we (Kubernetes SIG Network) now plan to remove it in Kubernetes 1.40. All ipvs users should therefore be planning to migrate to either the nftables or iptables backends. Why was the ipvs backend written? The ipvs backend was originally designed to fix two different kinds of performance problems in the iptables backend: First, iptables had problems with data plane latency (the speed at which clients could establish new connections to services), because kube-proxy would write a separate iptables rule to match each service IP, so every new connection would need to be matched against a long series of rules that was O(n) in the number of services in the cluster. Second, it had an even larger problem with control plane latency (the speed at which kube-proxy could process changes to services and endpoints and update the data plane rules), because kube-proxy would rewrite the entire iptables ruleset every time any service or endpoint changed. While the list of rules itself was only O(n) in the total number of endpoints,

## 40 Kubernetes Interview Questions That Expose Weak Senior Engineers

DevFeed: [40 Kubernetes Interview Questions That Expose Weak Senior Engineers](<https://devfeed.tech/articles/40-kubernetes-interview-questions-that-expose-weak-senior-engineers-136234.md>)

Original publisher: [Read original article](<https://blog.devops.dev/40-kubernetes-interview-questions-that-expose-weak-senior-engineers-f183baad39fb?source=rss----33f8b2d9a328---4>)

Author: Devrim Ozcay | Founder, ProdRescue

Published: 2026-11-16T13:26:05Z

Content type: article

Language: en

Sources: [DevOps.dev](<https://devfeed.tech/sources/devops-dev.md>)

Topics: [Production Engineering](<https://devfeed.tech/topics/production-engineering.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>), [control-plane](<https://devfeed.tech/topics/control-plane.md>), [consistency](<https://devfeed.tech/topics/consistency.md>), [zero downtime deployment](<https://devfeed.tech/topics/zero-downtime-deployment.md>), [health check](<https://devfeed.tech/topics/health-check.md>), [rate-limiting](<https://devfeed.tech/topics/rate-limiting.md>), [Access Control](<https://devfeed.tech/topics/access-control.md>), [Post Mortem](<https://devfeed.tech/topics/post-mortem.md>)

Tags: [advanced](<https://devfeed.tech/tags/advanced.md>), [affinity](<https://devfeed.tech/tags/affinity.md>), [autoscaling](<https://devfeed.tech/tags/autoscaling.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [devops](<https://devfeed.tech/tags/devops.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [docker](<https://devfeed.tech/tags/docker.md>), [failure](<https://devfeed.tech/tags/failure.md>), [interview](<https://devfeed.tech/tags/interview.md>), [interview-questions](<https://devfeed.tech/tags/interview-questions.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [kubernetes-cluster](<https://devfeed.tech/tags/kubernetes-cluster.md>), [latency](<https://devfeed.tech/tags/latency.md>), [load-balancing](<https://devfeed.tech/tags/load-balancing.md>), [memory](<https://devfeed.tech/tags/memory.md>), [networking](<https://devfeed.tech/tags/networking.md>), [persistent-volume](<https://devfeed.tech/tags/persistent-volume.md>), [production](<https://devfeed.tech/tags/production.md>), [programming](<https://devfeed.tech/tags/programming.md>), [questions](<https://devfeed.tech/tags/questions.md>), [routing](<https://devfeed.tech/tags/routing.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>), [staff](<https://devfeed.tech/tags/staff.md>)

### AI overview

The article argues that senior Kubernetes troubleshooting requires tracing how scheduling, networking, application lifecycle, resources, and dependencies interact. It explains practical failure patterns and diagnostics: a Pending Pod may lack an eligible node despite spare aggregate CPU; CPU limits can throttle latency-sensitive work; premature readiness can route traffic to unprepared instances; and autoscaling can worsen overload by increasing pressure on a saturated dependency. It recommends checking scheduling events and requests, probe behavior, EndpointSlices, DNS and conntrack metrics, resource throttling, traces, and downstream capacity. Other sections cover safe rollouts and shutdown, database connection budgets, stateful workload recovery, security boundaries, incident response, and tradeoffs in multi-zone, multi-region, and multi-tenant designs. The text presents hypothetical interview scenarios and diagnostic approaches, not evidence of a specific incident or a complete Kubernetes reference.

### Source excerpt

Advanced Kubernetes troubleshooting, distributed systems, networking, autoscaling, and production reliability for Senior and Staff Engineers. A Kubernetes cluster with 180 nodes had plenty of available CPU. Every deployment reported success, and the control plane appeared healthy. Yet customers were experiencing intermittent 503 errors, and the API's p99 latency had climbed from 180 milliseconds to almost nine seconds. The infrastructure team increased the number of replicas. The application team increased connection pool sizes. Someone suggested moving the workloads to larger nodes because the existing cluster was obviously struggling. The situation became worse. The problem was that newly created Pods were passing readiness checks before their applications could serve real traffic. Every additional replica introduced another temporarily unhealthy endpoint into the load-balancing path. Autoscaling was increasing the number of requests routed toward instances that were technically running but operationally useless. This is why Kubernetes interviews become considerably harder at the Staff level. Knowing which command describes a Pod is useful, but it does not demonstrate an understanding of how scheduling, network routing, application lifecycle, resource allocation, and distributed dependencies interact during failure. The following 40 questions focus on those interactions. The scenarios are hypothetical, but the failure mechanisms and diagnostic approaches are grounded in Kubernetes and production engineering behavior. Part I -- Kubernetes Scheduling and Resource Management1. A Pod is Pending even though the cluster has available CPU and memory. Why? A candidate who immediately recommends adding nodes is missing several possibilities. Kubernetes scheduling evaluates requested resources alongside node selectors, affinity rules, taints, tolerations, topology constraints, persistent-volume placement, and other scheduling restrictions. Aggregate cluster capacity is irrel

## Root Cause of Microsoft 365 Outage: July Bug Lay Dormant for Three Months as Test Environment Failed to Reproduce Production

DevFeed: [Root Cause of Microsoft 365 Outage: July Bug Lay Dormant for Three Months as Test Environment Failed to Reproduce Production](<https://devfeed.tech/articles/root-cause-of-microsoft-365-outage-july-bug-lay-dormant-for-three-months-as-test-environment-failed-to-reproduce-production-135302.md>)

Original publisher: [Read original article](<https://joho-todai.com/en/microsoft-365-outage-july-bug-three-month-latent/>)

Author: Toru Misaki

Published: 2026-11-16T10:29:10Z

Content type: news

Language: en

Sources: [Joho no Todai](<https://devfeed.tech/sources/joho-no-todai.md>)

Topics: [microsoft 365](<https://devfeed.tech/topics/microsoft-365.md>), [Microsoft](<https://devfeed.tech/topics/microsoft.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [staging](<https://devfeed.tech/topics/staging.md>), [blue green and canary deployments](<https://devfeed.tech/topics/blue-green-and-canary-deployments.md>)

Tags: [bug](<https://devfeed.tech/tags/bug.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [cause](<https://devfeed.tech/tags/cause.md>), [environment](<https://devfeed.tech/tags/environment.md>), [failed](<https://devfeed.tech/tags/failed.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-365](<https://devfeed.tech/tags/microsoft-365.md>), [microsoft-365-outage](<https://devfeed.tech/tags/microsoft-365-outage.md>), [outage](<https://devfeed.tech/tags/outage.md>), [production](<https://devfeed.tech/tags/production.md>), [root](<https://devfeed.tech/tags/root.md>), [test](<https://devfeed.tech/tags/test.md>)

### AI overview

A latent lock-contention defect in Microsoft 365 network gateway software caused the November 12 outage when a Teams failover increased load on the primary gateway while an unrelated issue had taken about 14% of server capacity offline. Microsoft's load tests used a different traffic mix and hardware configurations, weighted toward older, lower-core-count servers, so they did not reproduce conditions on newer production servers where the defect worsened. Traffic redirection then concentrated load on remaining gateways, and a separate, undisclosed issue prevented regional bulkheads from containing the cascade; over 80% of gateway servers were unresponsive at the peak. Teams, OneDrive, SharePoint, and Office apps recovered relatively quickly by moving off the affected infrastructure, while Exchange Online suffered longer because it could not. Normal recovery repeatedly overloaded restarted servers, and rollback was unavailable, so Microsoft applied an update and manually repaired servers in batches. The preliminary report leaves the bulkhead failure details and recurrence-prevention measures for a later report.

### Source excerpt

Microsoft has released a preliminary post-incident report on the major Microsoft 365 outage of November 12. A software defect introduced in July went unnoticed during more than three months of global deployment, then surfaced when increased load and reduced capacity converged; in under two hours it spread across the server fleet, leaving more than 80% of gateway servers unresponsive at the peak.

## PF-Exec: Large-Scale UAV Relay Load Balancing with an Anchored Potential Direction Decider and a Per-Dimension Gain-Learning Executor

DevFeed: [PF-Exec: Large-Scale UAV Relay Load Balancing with an Anchored Potential Direction Decider and a Per-Dimension Gain-Learning Executor](<https://devfeed.tech/articles/pf-exec-large-scale-uav-relay-load-balancing-with-an-anchored-potential-direction-decider-and-a-per-dimension-gain-learning-executor-135786.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.17492>)

Author: Liheng Zeng, Hanzhang Jia, Yiran Li, Yi Gao, Bo Ma

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [AI Networking](<https://devfeed.tech/topics/ai-networking.md>)

Tags: [balancing](<https://devfeed.tech/tags/balancing.md>), [cs-it](<https://devfeed.tech/tags/cs-it.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [dimension](<https://devfeed.tech/tags/dimension.md>), [eess-sp](<https://devfeed.tech/tags/eess-sp.md>), [executor](<https://devfeed.tech/tags/executor.md>), [large](<https://devfeed.tech/tags/large.md>), [learning](<https://devfeed.tech/tags/learning.md>), [load](<https://devfeed.tech/tags/load.md>), [load-balancing](<https://devfeed.tech/tags/load-balancing.md>), [math-it](<https://devfeed.tech/tags/math-it.md>), [relay](<https://devfeed.tech/tags/relay.md>)

### AI overview

PF-Exec separates large-scale UAV relay load balancing into an anchored potential direction decider and a gain-learning executor. The decider supplies closed-form targets; the executor learns a positive gain for each dimension from local observations and clips actions to physical single-step limits. Unit gains make its initial behavior equivalent to the pure potential policy. Bandwidth targets are closed form, while position targets use the demand-weighted centroid of served users. In the highest-pressure burst tests, the fixed-step comparison loses both fairness and mean load on some seeds; PF-Exec keeps fairness near one. Of ten comparison methods, only the greedy heuristic survives that setting, with much higher load variance. Fairness also stays high with stale global state and position noise in the two largest scenarios. Learned and fixed-step executor flight energy is nearly identical. The abstract reports proofs of convergence properties for both layers, but provides no detailed numerical results or proof assumptions.

### Source excerpt

arXiv:2611.17492v1 Announce Type: new Abstract: Relay load balancing in large-scale UAV networks requires joint decisions on association, bandwidth, power and hovering position, while the dimension grows with the number of UAVs and each UAV sees only its local state. This paper splits the decision into two layers. An anchored potential direction decider emits a closed-form target at every instant and returns its direction only. A gain-learning executor maps a local observation into a positive gain per dimension, composes gain and target, and clips the action against the physical single-step bound; the two layers form PF-Exec. The unit gain is the default start, so the executor reduces to the pure potential policy before training. The load is first-order homogeneous in bandwidth, making the bandwidth target closed form, and the position target is the demand-weighted centroid of the served users. The system uses a two-hop access and millimetre-wave backhaul link with soft association, a 3GPP UMa-AV channel with two shadowing layers, an OU-type social-force crowd and a quadrotor physical layer. Evaluation spans all fleet sizes, two crowd scenarios, independent seeds, the published baselines in their own settings and a degraded-information line. In the highest-pressure burst setting the fixed-step comparison breaks down on part of the seeds, losing fairness and mean load at once, while PF-Exec holds fairness near one. Among the ten comparison methods only the greedy heuristic survives that setting; its load variance stays far higher. Under stale global state and position noise the fairness index stays high at both largest-scale scenarios. Flight energy is nearly identical between the learned and the fixed-step executor, and three constants are fixed once across all settings. The convergence properties of both layers, from the truncated gain iteration to the perturbed tracking bound, are proved in the companion analysis.

## Fuse-All MoE Inference Kernel with Load Balancing on TPUs

DevFeed: [Fuse-All MoE Inference Kernel with Load Balancing on TPUs](<https://devfeed.tech/articles/fuse-all-moe-inference-kernel-with-load-balancing-on-tpus-135835.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.17857>)

Author: Rupeng Liu, Bangsheng Tang, Zongwei Zhou, Chengji Yao, Yejin Lee

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [balancing](<https://devfeed.tech/tags/balancing.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data](<https://devfeed.tech/tags/data.md>), [data-movement](<https://devfeed.tech/tags/data-movement.md>), [decode](<https://devfeed.tech/tags/decode.md>), [disaggregated-serving](<https://devfeed.tech/tags/disaggregated-serving.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [experts](<https://devfeed.tech/tags/experts.md>), [fuse](<https://devfeed.tech/tags/fuse.md>), [inference](<https://devfeed.tech/tags/inference.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [latency](<https://devfeed.tech/tags/latency.md>), [llms](<https://devfeed.tech/tags/llms.md>), [load](<https://devfeed.tech/tags/load.md>), [load-balancing](<https://devfeed.tech/tags/load-balancing.md>), [mixture-of-experts-moe](<https://devfeed.tech/tags/mixture-of-experts-moe.md>), [moe](<https://devfeed.tech/tags/moe.md>), [optimizations](<https://devfeed.tech/tags/optimizations.md>), [tpus](<https://devfeed.tech/tags/tpus.md>)

### AI overview

Fuse-All MoE targets TPU MoE inference bottlenecks from intermediate data movement and uneven expert loads by fusing token collection, expert projections, gated activation, and output writes while keeping computation intermediates on chip. Fixed token-slot destinations let scheduling and expert placement change without changing routing semantics. In disaggregated serving, it uses scheduling and expert redistribution to balance EP16 prefill workloads; during decode, symmetric EP8TP2 shares active expert computation across both TensorCore devices in each chip. For the measured skewed EP16 prefill workload, kernel runtime falls from 2.3x to 1.09x the fixed reference time for balanced workloads. Across 32-512 decode tokens, symmetric EP8TP2 lowers kernel latency by up to 21.3% versus fused EP16. With additional serving-stack optimizations, the integrated system reports up to 34.5% greater serving throughput than the Megablox EP16 baseline.

### Source excerpt

arXiv:2611.17857v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) architectures activate only a subset of experts for each token, allowing large language models (LLMs) to scale parameter capacity without a proportional increase in computation per token. However, intermediate data movement and uneven expert loads limit the efficiency of MoE inference on TPUs. To address these bottlenecks, we introduce Fuse-All MoE, a TPU kernel design that jointly optimizes expert computation, output communication, and load balancing. The kernel fuses token collection, expert projections, gated activation, and output writes, retaining expert computation intermediates on chip. Fixed token-slot destinations allow scheduling and expert placement to change while preserving routing semantics. In prefill-decode (PD) disaggregated serving, Fuse-All MoE balances EP16 prefill workloads through scheduling and expert redistribution, while symmetric EP8TP2 shares active expert computation between the two TensorCore devices within each chip during decode. Our experiments show that Fuse-All MoE reduces kernel runtime for a measured skewed EP16 prefill workload from 2.3x to 1.09x the same fixed reference time for a balanced workload. Across 32-512 decode tokens, symmetric EP8TP2 reduces kernel latency by up to 21.3% relative to fused EP16. With accompanying serving stack optimizations, the integrated system improves serving throughput by up to 34.5% over the Megablox EP16 baseline.

## Introducing Red Hat OpenShift Service Mesh 3.5

DevFeed: [Introducing Red Hat OpenShift Service Mesh 3.5](<https://devfeed.tech/articles/introducing-red-hat-openshift-service-mesh-3-5-135990.md>)

Original publisher: [Read original article](<https://www.redhat.com/en/blog/introducing-red-hat-openshift-service-mesh-35>)

Author: Jamie Longmuir

Published: 2026-11-16T00:00:00Z

Content type: release

Language: en

Sources: [Red Hat Blog](<https://devfeed.tech/sources/red-hat-blog.md>)

Topics: [service-mesh](<https://devfeed.tech/topics/service-mesh.md>), [istio](<https://devfeed.tech/topics/istio.md>), [SPIRE](<https://devfeed.tech/topics/spire.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Zero Trust](<https://devfeed.tech/topics/zero-trust.md>), [Deployment Strategies](<https://devfeed.tech/topics/deployment-strategies.md>), [control-plane](<https://devfeed.tech/topics/control-plane.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-chatbot](<https://devfeed.tech/tags/ai-chatbot.md>), [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [ambient](<https://devfeed.tech/tags/ambient.md>), [dual-stack](<https://devfeed.tech/tags/dual-stack.md>), [envoy](<https://devfeed.tech/tags/envoy.md>), [general-availability](<https://devfeed.tech/tags/general-availability.md>), [ipv4](<https://devfeed.tech/tags/ipv4.md>), [ipv6](<https://devfeed.tech/tags/ipv6.md>), [istio](<https://devfeed.tech/tags/istio.md>), [kiali](<https://devfeed.tech/tags/kiali.md>), [mesh](<https://devfeed.tech/tags/mesh.md>), [multicluster](<https://devfeed.tech/tags/multicluster.md>), [openshift](<https://devfeed.tech/tags/openshift.md>), [red-hat-openshift](<https://devfeed.tech/tags/red-hat-openshift.md>), [service-mesh](<https://devfeed.tech/tags/service-mesh.md>), [sidecar](<https://devfeed.tech/tags/sidecar.md>), [spire](<https://devfeed.tech/tags/spire.md>), [workload-identity](<https://devfeed.tech/tags/workload-identity.md>), [zero-trust](<https://devfeed.tech/tags/zero-trust.md>)

### AI overview

OpenShift Service Mesh 3.5 is generally available on OpenShift Container Platform 4.20 and later, including OpenShift 5.0. Based on Istio 1.31 and Kiali 2.33, it makes ambient-mode multiprimary multicluster and dual-stack IPv4/IPv6 generally available, and adds generally available SPIRE workload identity integration in sidecar mode, including federation; SPIRE is not yet supported in ambient mode. Sidecar-mode users can skip one Service Mesh minor release with revision-based upgrades, allowing workloads to jump two Service Mesh minor releases with one restart. Highlights include inherited mesh-wide resilience defaults, zone-aware load balancing, secure metrics scraping, and dynamic DNS for external HTTP destinations. Canary waypoint upgrades and standalone VM integration are Technology Preview; fleet management through Advanced Cluster Management is Developer Preview. Ambient multicluster currently supports multiprimary, multinetwork topology, while other topologies remain unsupported.

### Source excerpt

Red Hat OpenShift Service Mesh 3.5 is now generally available with Red Hat OpenShift Container Platform. Based on the Istio, Envoy, and Kiali projects, this release updates the version of Istio to 1.31 and Kiali to 2.33, and is supported on OpenShift Container Platform 4.20 and above, including Red Hat OpenShift 5.0. This release graduates multicluster with ambient mode, dual-stack IPv4/IPv6 with ambient mode, and Red Hat's zero trust workload identity manager (SPIRE) with sidecar mode to general availability, along with a substantial set of enhancements from Istio 1.31. Red Hat OpenShift 5.0

## Fastly Recognized in the Inaugural Gartner® Magic Quadrant™ for Edge Distribution Platforms

DevFeed: [Fastly Recognized in the Inaugural Gartner® Magic Quadrant™ for Edge Distribution Platforms](<https://devfeed.tech/articles/fastly-recognized-in-the-inaugural-gartner-magic-quadranttm-for-edge-distribution-platforms-130635.md>)

Original publisher: [Read original article](<https://www.fastly.com/blog/fastly-recognized-gartner-magic-quadrant-edge-distribution-platforms/>)

Author: Skip Bacon

Published: 2026-11-12T00:00:00Z

Content type: release

Language: en

Sources: [Fastly Blog](<https://devfeed.tech/sources/fastly-blog.md>)

Topics: [Cloudflare Workers](<https://devfeed.tech/topics/cloudflare-workers.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Internet Traffic](<https://devfeed.tech/topics/internet-traffic.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-services](<https://devfeed.tech/tags/ai-services.md>), [api](<https://devfeed.tech/tags/api.md>), [api-governance](<https://devfeed.tech/tags/api-governance.md>), [apis](<https://devfeed.tech/tags/apis.md>), [applications](<https://devfeed.tech/tags/applications.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [bot-management](<https://devfeed.tech/tags/bot-management.md>), [cache](<https://devfeed.tech/tags/cache.md>), [cache-invalidation](<https://devfeed.tech/tags/cache-invalidation.md>), [caching](<https://devfeed.tech/tags/caching.md>), [commerce](<https://devfeed.tech/tags/commerce.md>), [company-news](<https://devfeed.tech/tags/company-news.md>), [compute](<https://devfeed.tech/tags/compute.md>), [ddos](<https://devfeed.tech/tags/ddos.md>), [edge](<https://devfeed.tech/tags/edge.md>), [fastly](<https://devfeed.tech/tags/fastly.md>), [gartner](<https://devfeed.tech/tags/gartner.md>), [gartner-magic-quadranttm](<https://devfeed.tech/tags/gartner-magic-quadranttm.md>), [leader](<https://devfeed.tech/tags/leader.md>)

### AI overview

Fastly says Gartner named it a Leader in the inaugural 2026 Magic Quadrant for Edge Distribution Platforms. The company describes edge use as extending beyond caching to application logic, API protection, real-time traffic monitoring, and emerging AI traffic governance. Its stated approach combines WebAssembly-based serverless compute, support for Rust, Go, JavaScript, and C++, and fast cache invalidation with intelligent routing intended to steer around network degradation. Fastly also says it integrates API protection, bot management, DDoS defense, and real-time observability with delivery and compute. Planned investments include edge AI services, active API governance, and natural-language platform management. The excerpt presents Fastly's own claims and notes Gartner's research is opinion, not endorsement.

### Source excerpt

Fastly has been named a Leader in the inaugural 2026 Gartner® Magic Quadrant™ for Edge Distribution Platforms.

## API Gateway vs Load Balancer: Do You Need Both?

DevFeed: [API Gateway vs Load Balancer: Do You Need Both?](<https://devfeed.tech/articles/api-gateway-vs-load-balancer-do-you-need-both-129367.md>)

Original publisher: [Read original article](<https://blog.levelupcoding.com/p/api-gateway-vs-load-balancer>)

Author: Nikki Siapno

Published: 2026-11-11T11:49:28Z

Content type: comparison

Language: en

Sources: [Level Up Coding System Design Newsletter](<https://devfeed.tech/sources/level-up-coding-system-design-newsletter.md>)

Topics: [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [API](<https://devfeed.tech/topics/api.md>), [health check](<https://devfeed.tech/topics/health-check.md>), [rate-limiting](<https://devfeed.tech/topics/rate-limiting.md>), [API Governance](<https://devfeed.tech/topics/api-governance.md>), [API keys](<https://devfeed.tech/topics/api-keys.md>)

Tags: [algorithms](<https://devfeed.tech/tags/algorithms.md>), [api](<https://devfeed.tech/tags/api.md>), [api-gateway](<https://devfeed.tech/tags/api-gateway.md>), [api-management](<https://devfeed.tech/tags/api-management.md>), [apis](<https://devfeed.tech/tags/apis.md>), [application](<https://devfeed.tech/tags/application.md>), [application-load-balancer](<https://devfeed.tech/tags/application-load-balancer.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [availability](<https://devfeed.tech/tags/availability.md>), [aws](<https://devfeed.tech/tags/aws.md>), [backend](<https://devfeed.tech/tags/backend.md>), [backends](<https://devfeed.tech/tags/backends.md>), [balancing](<https://devfeed.tech/tags/balancing.md>), [checkout](<https://devfeed.tech/tags/checkout.md>), [e-commerce](<https://devfeed.tech/tags/e-commerce.md>), [load-balancer](<https://devfeed.tech/tags/load-balancer.md>)

### AI overview

A load balancer distributes traffic among healthy backend instances; an API gateway manages client access through policies such as authentication, quotas, transformations, and API versioning, then routes requests to services. Their features can overlap: AWS Application Load Balancer supports some authentication and URL rewriting, while Kong offers balancing and health checks. Choose a load balancer for server distribution or Layer 4 non-HTTP traffic, a gateway for centralized API policies, and both when those jobs are distinct or multiple self-hosted gateways need balancing. A single product may cover both at the required scale. Gateways do not replace services' own authorization checks.

### Source excerpt

Modern API gateways can load balance. Modern load balancers can handle authentication. So why use both?

## Reconciling Strict SLOs with Low Cost under Fine-Grained Load Fluctuations

DevFeed: [Reconciling Strict SLOs with Low Cost under Fine-Grained Load Fluctuations](<https://devfeed.tech/articles/reconciling-strict-slos-with-low-cost-under-fine-grained-load-fluctuations-127790.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.30533>)

Author: Dilina Dehigama, Shyam Jesalpura, Zeyu Xu, Marton Nemeth, Shengda Zhu, Marios Kogias, Boris Grot

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [function as a service](<https://devfeed.tech/topics/function-as-a-service.md>), [virtual machines](<https://devfeed.tech/topics/virtual-machines.md>), [cloud cost management](<https://devfeed.tech/topics/cloud-cost-management.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [hosting](<https://devfeed.tech/topics/hosting.md>)

Tags: [accounting](<https://devfeed.tech/tags/accounting.md>), [application](<https://devfeed.tech/tags/application.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [autoscalers](<https://devfeed.tech/tags/autoscalers.md>), [aws](<https://devfeed.tech/tags/aws.md>), [aws-lambda](<https://devfeed.tech/tags/aws-lambda.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-based](<https://devfeed.tech/tags/cloud-based.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [compliance](<https://devfeed.tech/tags/compliance.md>), [controller](<https://devfeed.tech/tags/controller.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cost-efficiency](<https://devfeed.tech/tags/cost-efficiency.md>), [cost-optimization](<https://devfeed.tech/tags/cost-optimization.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cpu-utilization](<https://devfeed.tech/tags/cpu-utilization.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [decoupling](<https://devfeed.tech/tags/decoupling.md>), [deployments](<https://devfeed.tech/tags/deployments.md>), [load](<https://devfeed.tech/tags/load.md>), [slos](<https://devfeed.tech/tags/slos.md>)

### AI overview

Spandana separates SLO enforcement from cost optimization for services facing sub-second load swings. A lightweight controller on each application VM keeps requests there when they can meet the SLO and sends the rest to a stock FaaS layer such as AWS Lambda. A resource allocator chooses VM provisioning based on VM and FaaS costs and traffic volatility. The reported evaluation finds strict SLO adherence, 76-86% CPU utilization, and 5-44% lower cost than three state-of-the-art baselines. The supplied text does not provide details about the evaluation setup or individual baselines.

### Source excerpt

arXiv:2606.30533v2 Announce Type: replace Abstract: Cloud-based online services face significant sub-second load fluctuations while needing to meet strict Service Level Objectives (SLOs). Cluster operators often over-provision resources to protect SLOs, sacrificing utilization and cost efficiency. Existing reactive and proactive autoscalers, serverless (FaaS) deployments, and VM/FaaS hybrid systems fail to reconcile strict SLO compliance with low cost and high utilization under fine-grained load fluctuation. We introduce Spandana, an architecture that addresses this trade off by decoupling SLO enforcement from cost optimization. A lightweight controller colocated with each application VM enforces SLOs by steering each arriving request between the VM and FaaS. Requests that can meet the SLO stay on the VM; the remaining requests are forwarded to a stock FaaS layer such as AWS Lambda. For cost optimization, Spandana's resource allocator determines the most-efficient VM provisioning by accounting for VM cost, FaaS cost, and traffic volatility, allowing the VM pool to run at high utilization. Our evaluation shows that Spandana maintains strict SLO adherence, achieves 76-86% CPU utilization, and reduces cost by 5-44% over three SOTA baselines.

## Meta Skipped Its Own Deploy Freeze for the World Cup Final.

DevFeed: [Meta Skipped Its Own Deploy Freeze for the World Cup Final.](<https://devfeed.tech/articles/meta-skipped-its-own-deploy-freeze-for-the-world-cup-final-127573.md>)

Original publisher: [Read original article](<https://blog.devops.dev/meta-skipped-its-own-deploy-freeze-for-the-world-cup-final-4d61566792b2?source=rss----33f8b2d9a328---4>)

Author: Vortex 404

Published: 2026-11-10T09:43:01Z

Content type: news

Language: en

Sources: [DevOps.dev](<https://devfeed.tech/sources/devops-dev.md>)

Topics: [Disaster Recovery](<https://devfeed.tech/topics/disaster-recovery.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [argentina](<https://devfeed.tech/tags/argentina.md>), [code](<https://devfeed.tech/tags/code.md>), [cup](<https://devfeed.tech/tags/cup.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [devops](<https://devfeed.tech/tags/devops.md>), [disaster-recovery](<https://devfeed.tech/tags/disaster-recovery.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [event](<https://devfeed.tech/tags/event.md>), [freeze](<https://devfeed.tech/tags/freeze.md>), [history](<https://devfeed.tech/tags/history.md>), [load](<https://devfeed.tech/tags/load.md>), [meta](<https://devfeed.tech/tags/meta.md>), [outages](<https://devfeed.tech/tags/outages.md>), [platform](<https://devfeed.tech/tags/platform.md>), [regions](<https://devfeed.tech/tags/regions.md>), [spain](<https://devfeed.tech/tags/spain.md>), [sre](<https://devfeed.tech/tags/sre.md>), [traffic](<https://devfeed.tech/tags/traffic.md>), [whatsapp](<https://devfeed.tech/tags/whatsapp.md>), [world](<https://devfeed.tech/tags/world.md>), [zero](<https://devfeed.tech/tags/zero.md>)

### AI overview

During the July 19, 2026 Argentina-Spain World Cup final, WhatsApp delivered more than 30 million messages per second, 1.23 times its previous all-time peak. The excerpt says Meta kept its usual engineering cadence across 48 teams and 104 matches, without shifting load between regions, imposing a code freeze, or pausing disaster-recovery drills. Meta reported zero outages or degradations. The excerpt does not include the engineering details behind those results.

### Source excerpt

No regional load-shifting, no code freeze, no paused disaster-recovery drills -- during the biggest traffic event in company history, Meta... Continue reading on DevOps.dev "

## How Do You Design a Highly Available Architecture in AWS? A Complete DevOps Engineer's Guide

DevFeed: [How Do You Design a Highly Available Architecture in AWS? A Complete DevOps Engineer's Guide](<https://devfeed.tech/articles/how-do-you-design-a-highly-available-architecture-in-aws-a-complete-devops-engineer-s-guide-127571.md>)

Original publisher: [Read original article](<https://blog.devops.dev/how-do-you-design-a-highly-available-architecture-in-aws-a-complete-devops-engineers-guide-fe49b3cc18d8?source=rss----33f8b2d9a328---4>)

Author: Arvind Verma

Published: 2026-11-10T09:42:59Z

Content type: article

Language: en

Sources: [DevOps.dev](<https://devfeed.tech/sources/devops-dev.md>)

Topics: [Cloud Architecture](<https://devfeed.tech/topics/cloud-architecture.md>), [software-architecture](<https://devfeed.tech/topics/software-architecture.md>), [reliability-engineering](<https://devfeed.tech/topics/reliability-engineering.md>), [system-design-project](<https://devfeed.tech/topics/system-design-project.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Amazon Route 53](<https://devfeed.tech/topics/amazon-route-53.md>), [Amazon VPC](<https://devfeed.tech/topics/amazon-vpc.md>), [DevOps](<https://devfeed.tech/topics/devops.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [auto-scaling](<https://devfeed.tech/tags/auto-scaling.md>), [availability](<https://devfeed.tech/tags/availability.md>), [aws](<https://devfeed.tech/tags/aws.md>), [cloudfront](<https://devfeed.tech/tags/cloudfront.md>), [database](<https://devfeed.tech/tags/database.md>), [design](<https://devfeed.tech/tags/design.md>), [designing](<https://devfeed.tech/tags/designing.md>), [devops](<https://devfeed.tech/tags/devops.md>), [devops-engineer](<https://devfeed.tech/tags/devops-engineer.md>), [disaster-recovery](<https://devfeed.tech/tags/disaster-recovery.md>), [docker](<https://devfeed.tech/tags/docker.md>), [e-commerce](<https://devfeed.tech/tags/e-commerce.md>), [ec2](<https://devfeed.tech/tags/ec2.md>), [high-availability](<https://devfeed.tech/tags/high-availability.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [implementation](<https://devfeed.tech/tags/implementation.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [interview](<https://devfeed.tech/tags/interview.md>), [interview-questions](<https://devfeed.tech/tags/interview-questions.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [load-balancers](<https://devfeed.tech/tags/load-balancers.md>)

### AI overview

A highly available AWS design spreads application components across multiple Availability Zones so a single component or AZ failure does not take down the service. The guide's example uses Route 53 for DNS, WAF for request filtering, an internet facing ALB to route traffic to healthy EC2 targets in private subnets, and an Auto Scaling Group to replace unhealthy instances and adjust capacity. It pairs RDS Multi-AZ failover with application retry and reconnection logic, while noting that a traditional standby is not for read scaling. S3 and CloudFront serve static assets; CloudWatch alarms and SNS notify operators. The guide distinguishes regional high availability from disaster recovery, recommending backups, recovery objectives, cross-Region plans where needed, and tested recovery. It also flags costs, DNS caching delays, NAT Gateway failure dependencies, and the need to test actual failure behavior.

### Source excerpt

From Route 53 and Application Load Balancers to Multi-AZ Auto Scaling, RDS, CloudFront, and Disaster Recovery. Understand how to design AWS infrastructure that survives failures. Imagine you're working as a DevOps Engineer at an e-commerce company. Your application handles thousands of customer requests every minute. Customers are browsing products, placing orders, making payments, and checking delivery status. Everything is working perfectly. Suddenly, one AWS Availability Zone experiences an infrastructure failure. Your EC2 instances become unavailable. Some application requests start failing. Customers cannot access your website. Your engineering manager asks: "How did our entire application become unavailable because one Availability Zone failed?" This is where understanding High Availability (HA) becomes important. A production-ready architecture should not depend on a single EC2 instance, one Availability Zone, one database instance, or one network component. Instead, infrastructure should be designed to tolerate failures, recover automatically, and maintain application availability. In this article, we'll understand how to design a highly available AWS architecture from scratch, including the architecture, AWS services, implementation decisions, production scenarios, and common interview questions. 📘 SECTION 1: UNDERSTANDING HIGH AVAILABILITY AND AWS ARCHITECTURE1. What Is High Availability? High Availability means designing systems that remain accessible and operational even when individual infrastructure components fail. Let's understand with a simple example. Imagine your application runs on one EC2 instance. If that instance crashes, your application becomes unavailable. This is called a Single Point of Failure (SPOF). Now imagine running six EC2 instances across two Availability Zones. Availability Zone A contains three instances. Availability Zone B contains three instances. If Availability Zone A experiences a failure, instances in Availability Zone B

## A Per-Step Load-Balancing Engine for Expert-Parallel MoE Serving

DevFeed: [A Per-Step Load-Balancing Engine for Expert-Parallel MoE Serving](<https://devfeed.tech/articles/a-per-step-load-balancing-engine-for-expert-parallel-moe-serving-126726.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.11402>)

Author: Jingyuan Li, Maolin Wang, Hayden So

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>), [GPU optimization](<https://devfeed.tech/topics/gpu-optimization.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>)

Tags: [balancing](<https://devfeed.tech/tags/balancing.md>), [cs-ar](<https://devfeed.tech/tags/cs-ar.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [engine](<https://devfeed.tech/tags/engine.md>), [load](<https://devfeed.tech/tags/load.md>), [load-balancing](<https://devfeed.tech/tags/load-balancing.md>), [moe](<https://devfeed.tech/tags/moe.md>), [parallel](<https://devfeed.tech/tags/parallel.md>), [serving](<https://devfeed.tech/tags/serving.md>)

### AI overview

Relief balances expert-parallel MoE workloads anew at each decode layer and step, dividing tokens across replicated expert copies to reduce GPU-group stragglers. Its bidirectional pairwise relaxation replaces exact linear-program solving with a few passes of integer halving over token counts, approaching the per-step optimum while avoiding solver latency and working-set limits described for deployment-scale GPU shared memory. The paper measures CUDA kernels on the serving GPU, an FPGA over PCIe, and a fixed-function block beside the accelerator. Across decode routing from three open MoE families, the abstract reports near-optimal balance at measured group sizes and layer times; the kernel matches or exceeds deployable schemes from 100-microsecond layers onward, while the on-die engine recovers up to 37.3 percentage points over deployed static placement at a stated transistor cost of 0.02% of a current accelerator.

### Source excerpt

arXiv:2611.11402v1 Announce Type: cross Abstract: In expert-parallel Mixture-of-Experts serving, a layer finishes only when the most loaded GPU of its group does. Token routing shifts at every decode step and concentrates on a few experts, so most of the group's reserved GPU time waits on stragglers. Recovering that time means re-deciding, at every layer of every decode step, how each replicated expert's tokens divide between its copies. The exact linear-program solvers that make this decision optimally take longer than many of the layers they serve, and their working set outgrows GPU shared memory at deployment scale. We introduce Relief, a bidirectional pairwise relaxation that makes the linear program unnecessary, coming within a fraction of a point of the per-step optimum in a few passes of integer halving over the token counts. We build and measure three deployments, CUDA kernels on the serving GPU, the same verified engine on an FPGA over PCIe, and a fixed-function block beside the accelerator. On decode routing from three open MoE families, Relief holds the balance near the per-step optimum at every group size and layer time measured. Scored as effective utilization relative to that optimum, the kernel alone matches or exceeds every deployable scheme at every group size, from the hundred-microsecond layer on. The on-die engine widens that lead at every point, recovering up to 37.3 percentage points over the deployed static placement, while adding 0.02% of a current accelerator's transistors.

## ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving

DevFeed: [ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving](<https://devfeed.tech/articles/eldr-expert-locality-aware-decode-routing-for-pd-disaggregated-moe-serving-123663.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.00466>)

Author: Sangjin Choi, Sukmin Cho, Yifan Xiong, Ziyue Yang, Youngjin Kwon, Peng Cheng

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [LLM Techniques](<https://devfeed.tech/topics/llm-techniques.md>), [vllm](<https://devfeed.tech/topics/vllm.md>), [Multi-GPU](<https://devfeed.tech/topics/multi-gpu.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [background](<https://devfeed.tech/tags/background.md>), [balancing](<https://devfeed.tech/tags/balancing.md>), [band](<https://devfeed.tech/tags/band.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [best](<https://devfeed.tech/tags/best.md>), [block](<https://devfeed.tech/tags/block.md>), [cache](<https://devfeed.tech/tags/cache.md>), [caching](<https://devfeed.tech/tags/caching.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [decode](<https://devfeed.tech/tags/decode.md>), [decoder](<https://devfeed.tech/tags/decoder.md>), [experts](<https://devfeed.tech/tags/experts.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [locality](<https://devfeed.tech/tags/locality.md>), [moe](<https://devfeed.tech/tags/moe.md>), [pd](<https://devfeed.tech/tags/pd.md>), [routing](<https://devfeed.tech/tags/routing.md>), [serving](<https://devfeed.tech/tags/serving.md>)

### AI overview

ELDR reduces decode latency in bandwidth-bound, PD-disaggregated MoE serving by routing requests toward workers likely to have the needed experts resident, while balancing load. It uses prefill signatures to predict decode expert usage, balanced K-means centroids refit in the background, locality-band routing to choose the least-loaded worker among close matches, and a block-granular cache to keep signatures exact under prefix caching. In vLLM evaluations on up to 40 GPUs, across three MoE models, 12 public task benchmarks, and multilingual WildChat, ELDR lowered median time per output token by 5.4-14.5% on average across request rates and by up to 17.8% versus the best of four load-balancing baselines. The reported gains preserved model outputs and token-to-expert assignments.

### Source excerpt

arXiv:2607.00466v3 Announce Type: replace Abstract: Equal load does not mean equal decode latency in prefill-decode (PD) disaggregated mixture-of-experts (MoE) serving. In bandwidth-bound decode, more distinct experts mean more weight loads and higher latency. We present ELDR, a lossless, expert-locality-aware decode router. Prefill signatures predict decode expert usage. Balanced $K$-means fits per-decoder centroids with background refitting. Locality-band routing selects the least-loaded worker among close matches. A block-granular cache keeps signatures exact under prefix caching. We evaluate ELDR in vLLM on up to 40 GPUs. Across three MoE models on 12 public task benchmarks and multilingual WildChat, it reduces median time-per-output-token (TPOT) by 5.4-14.5% on average across request rates and up to 17.8% over the best of four load-balancing baselines, preserving model outputs and token-to-expert assignments.

## Load Balancing with Item Fragmentation and More Bins than Items: Linear-Time and Exact

DevFeed: [Load Balancing with Item Fragmentation and More Bins than Items: Linear-Time and Exact](<https://devfeed.tech/articles/load-balancing-with-item-fragmentation-and-more-bins-than-items-linear-time-and-exact-123691.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.09267>)

Author: Ben Bals, Tom\\'{a}\\v{s} Flouri, Kassian Kobert, Solon P. Pissis, Leen Stougie

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>), [Combinatorial optimization](<https://devfeed.tech/topics/combinatorial-optimization.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>), [Concurrency](<https://devfeed.tech/topics/concurrency.md>)

Tags: [algorithms](<https://devfeed.tech/tags/algorithms.md>), [approximation](<https://devfeed.tech/tags/approximation.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [cs-ds](<https://devfeed.tech/tags/cs-ds.md>), [distributed-systems](<https://devfeed.tech/tags/distributed-systems.md>), [load-balancing](<https://devfeed.tech/tags/load-balancing.md>)

### AI overview

For load balancing with fragmentable integer-size items and bins whose loads differ by at most one, the goal is to minimize the maximum number of item fragments in any bin. In the case B >= N, the paper gives an O(N)-time, O(N)-space algorithm that finds an optimal assignment. It proves an optimum exists with at most two fragments per bin and characterizes when one fragment per bin is sufficient. Combined with the prior algorithm, the result yields a strongly polynomial-time (OPT+1)-approximation for the general case, removing dependence on the number of bins B.

### Source excerpt

arXiv:2611.09267v1 Announce Type: cross Abstract: We study a fundamental load-balancing problem in which $N$ items of strictly positive integer sizes with total size $W$ must be assigned to $B$ bins, with items allowed to be divided into integer-size fragments across bins. Each bin must receive a load of either $\lfloor W/B \rfloor$ or $\lceil W/B \rceil$, ensuring that the loads of any two bins differ by at most one unit. Under this constraint, the objective is to minimize the maximum number of distinct item fragments placed into any single bin. This problem is NP-hard and there exists an $(\mathrm{OPT}+1)$-approximation algorithm that runs in $O(N \log N + B)$ time [Kobert et al., 2014]. The existing algorithm is employed by several widely used production codes for phylogenetic reconstruction. Here, we focus on the case $B \ge N$, that is, bins are at least as numerous as items. This setting arises pervasively in modern parallel and distributed systems, as well as in iterative algorithms that rebalance their workload at every step. In this context, bins are machines or machine cores, and items represent computational tasks (jobs). Our central contribution is a remarkably simple $O(N)$-time algorithm that computes an optimal assignment in this setting; this assignment is encoded using $O(N)$ space. In particular, we prove that an optimum always exists in which every bin receives at most two fragments and provide a characterization of when one fragment per bin suffices. Our result strictly improves the $(\mathrm{OPT}+1)$-approximation of Kobert et al. in the many-bins setting and, together with it, yields a strongly polynomial-time $(\mathrm{OPT}+1)$-approximation algorithm for the general setting, as it removes the dependency on the number $B$ of bins.

[Next page](<https://devfeed.tech/topics/load-balancing.md?cursor=WyIyMDI2LTExLTA5VDA1OjAwOjAwKzAwOjAwIiwgImNlMDA4NDI1LTQyMTUtNDg2ZC04NDljLTIzMTk3ZjA4OGY2OCJd>)