# quasi-random network topology

A data-center network topology with fixed and random components that enables practical cabling of flat networks.

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

## Bridge: Optimizing Collective Communication Schedules in Reconfigurable Networks with Reusable Subrings

DevFeed: [Bridge: Optimizing Collective Communication Schedules in Reconfigurable Networks with Reusable Subrings](<https://devfeed.tech/articles/bridge-optimizing-collective-communication-schedules-in-reconfigurable-networks-with-reusable-subrings-127832.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.12766>)

Author: Anton Juerss, Stefan Schmid

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

Content type: article

Language: en

Sources: [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [quasi-random network topology](<https://devfeed.tech/topics/quasi-random-network-topology.md>)

Tags: [algorithm](<https://devfeed.tech/tags/algorithm.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [bridge](<https://devfeed.tech/tags/bridge.md>), [collective](<https://devfeed.tech/tags/collective.md>), [communication](<https://devfeed.tech/tags/communication.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [delay](<https://devfeed.tech/tags/delay.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [exploits](<https://devfeed.tech/tags/exploits.md>), [future](<https://devfeed.tech/tags/future.md>), [links](<https://devfeed.tech/tags/links.md>), [networks](<https://devfeed.tech/tags/networks.md>), [optical](<https://devfeed.tech/tags/optical.md>), [optimizing](<https://devfeed.tech/tags/optimizing.md>), [ring](<https://devfeed.tech/tags/ring.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [schedules](<https://devfeed.tech/tags/schedules.md>), [strategies](<https://devfeed.tech/tags/strategies.md>), [strategy](<https://devfeed.tech/tags/strategy.md>), [structure](<https://devfeed.tech/tags/structure.md>), [topology](<https://devfeed.tech/tags/topology.md>)

### AI overview

Bridge uses sparse optical network reconfiguration for All-to-All, AllReduce, Reduce-Scatter, and AllGather. It connects immediate communication partners to reduce propagation and transmission delay, while connected subrings preserve reachability to future peers so links can be reused across multiple steps and reconfiguration cost can be amortized. The abstract reports typically 3x-10x lower All-to-All completion time than static baselines, including effectiveness with millisecond-scale reconfiguration delays. For AllReduce, it reports up to 1.5x improvement over existing reconfiguration strategies and 1.5x-6.6x over Ring on low- to moderate-sized workloads.

### Source excerpt

arXiv:2605.12766v2 Announce Type: replace Abstract: Optical circuit-switched networks have emerged as an appealing alternative to electrical fabrics as they can reconfigure the network topology at runtime, reducing communication cost and improving bandwidth utilization. Yet exploiting optical reconfigurable networks for collective communication comes with a fundamental trade-off: each reconfiguration incurs non-negligible delay and the benefit of a new topology depends on future traffic. The central question is therefore when reconfiguration is worth its cost. While prior work has demonstrated the benefits of reconfiguration, the opportunity to reuse optical links to optimize multiple steps remains largely unexplored. In this paper, we present Bridge, a reconfiguration strategy for frequently used collective communication primitives, namely All-to-All, AllReduce, Reduce-Scatter, and AllGather. Bridge exploits the structure of Bruck's communication pattern to support efficient sparse reconfiguration. The key idea is to reduce propagation and transmission delay by directly connecting immediate communication partners and preserve efficient reachability to future peers through connected subrings. As a result, optical links can be reused across multiple subsequent steps, allowing the benefit of reconfiguration to amortize beyond a single step. Our evaluation shows that Bridge reduces All-to-All completion time by typically 3x to 10x over static baselines and is effective even with millisecond-scale reconfiguration delays; for AllReduce up to 1.5x against existing reconfiguration strategies, and by 1.5x to 6.6x against the Ring algorithm on low to moderate-sized workloads.

## Celestica DS6005 Puts 102.4Tbps of Co-Packaged Optics and 64 1.6TbE Ports in a 1.5OU Liquid-Cooled ORv3 Switch

DevFeed: [Celestica DS6005 Puts 102.4Tbps of Co-Packaged Optics and 64 1.6TbE Ports in a 1.5OU Liquid-Cooled ORv3 Switch](<https://devfeed.tech/articles/celestica-ds6005-puts-102-4tbps-of-co-packaged-optics-and-64-1-6tbe-ports-in-a-1-5ou-liquid-cooled-orv3-switch-88005.md>)

Original publisher: [Read original article](<https://www.storagereview.com/news/celestica-ds6005-102-4tbps-co-packaged-optics-1-5ou-liquid-cooled-orv3-switch>)

Author: Harold Fritts

Published: 2026-10-15T16:50:02Z

Content type: news

Language: en

Sources: [StorageReview.com](<https://devfeed.tech/sources/storagereview-com.md>)

Topics: [quasi-random network topology](<https://devfeed.tech/topics/quasi-random-network-topology.md>), [AWS data center network architecture](<https://devfeed.tech/topics/aws-data-center-network-architecture.md>), [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>)

Tags: [102-4t](<https://devfeed.tech/tags/102-4t.md>), [amd](<https://devfeed.tech/tags/amd.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [co-packaged-optics](<https://devfeed.tech/tags/co-packaged-optics.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [hardware-root-of-trust](<https://devfeed.tech/tags/hardware-root-of-trust.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [networking](<https://devfeed.tech/tags/networking.md>), [ocp](<https://devfeed.tech/tags/ocp.md>), [optics](<https://devfeed.tech/tags/optics.md>), [switch](<https://devfeed.tech/tags/switch.md>), [v3](<https://devfeed.tech/tags/v3.md>)

### AI overview

Celestica introduced the DS6005, a 102.4 Tbps co-packaged optics data center switch for Open Rack v3, with 64 optically enabled 1.6 TbE ports and liquid cooling. It uses Broadcom's Tomahawk 6 Davisson package, supports field-replaceable external laser modules, and runs SONiC. Celestica has not published power or latency figures or availability dates.

### Source excerpt

Celestica introduced the DS6005 at the OCP Global Summit, a co-packaged optics (CPO) data center switch that puts 102.4Tbps of switching capacity and 64 optically enabled 1.6TbE ports in a 1.5OU, fully liquid-cooled chassis for Open Rack v3. It is built on Broadcom's Tomahawk 6 Davisson, the CPO version of the 102.4T Tomahawk 6, which The post Celestica DS6005 Puts 102.4Tbps of Co-Packaged Optics and 64 1.6TbE Ports in a 1.5OU Liquid-Cooled ORv3 Switch appeared first on StorageReview.com.

## Geo-Aware Leader Schedule

DevFeed: [Geo-Aware Leader Schedule](<https://devfeed.tech/articles/geo-aware-leader-schedule-87551.md>)

Original publisher: [Read original article](<https://solana.com/news/geo-aware-leader-schedule>)

Author: Umberto Natale

Published: 2026-10-14T15:05:00Z

Content type: article

Language: en

Sources: [Solana News Feed](<https://devfeed.tech/sources/solana-news-feed.md>)

Topics: [quasi-random network topology](<https://devfeed.tech/topics/quasi-random-network-topology.md>), [Spotify](<https://devfeed.tech/topics/spotify.md>), [proof of authority](<https://devfeed.tech/topics/proof-of-authority.md>), [Service-Level Agreements](<https://devfeed.tech/topics/service-level-agreements.md>)

Tags: [algorithm](<https://devfeed.tech/tags/algorithm.md>), [blockchain](<https://devfeed.tech/tags/blockchain.md>), [blockchain-technology](<https://devfeed.tech/tags/blockchain-technology.md>), [crypto-news](<https://devfeed.tech/tags/crypto-news.md>), [cryptocurrency](<https://devfeed.tech/tags/cryptocurrency.md>), [defi](<https://devfeed.tech/tags/defi.md>), [latency](<https://devfeed.tech/tags/latency.md>), [network](<https://devfeed.tech/tags/network.md>), [networks](<https://devfeed.tech/tags/networks.md>), [nfts](<https://devfeed.tech/tags/nfts.md>), [podcasts](<https://devfeed.tech/tags/podcasts.md>), [random](<https://devfeed.tech/tags/random.md>), [solana](<https://devfeed.tech/tags/solana.md>), [solana-ecosystem](<https://devfeed.tech/tags/solana-ecosystem.md>), [upgrades](<https://devfeed.tech/tags/upgrades.md>), [web3](<https://devfeed.tech/tags/web3.md>)

### AI overview

This study assesses Solana's proposed geo-aware leader schedule, which reorders stake-weighted leader windows so consecutive validators are geographically closer, reducing estimated handoff latency without changing each validator's number of windows. Simulations suggest a stake floor near 10% and a run length between 3 and 7 balance latency gains against regional concentration and fictitious-location risks. The recommendations depend on the current set of validator locations and stake distribution.

### Source excerpt

This study evaluates SIMD-0675, a geo-aware leader schedule for Solana. It reorders the stake-weighted random schedule so that consecutive leaders are geographically close, which cuts handoff latency. No validator's number of leader windows changes. Two parameters are tuned: STAKE_FLOOR (σ), the stake fraction that sets each location's compatibility radius, and RUN_LENGTH (γ), the size of the contiguous blocks of compatible leaders. They are tested on 6,250 synthetic Monte Carlo networks built on today's 124 validator locations. The handoff gain levels off quickly as γ grows, while the risks keep rising: leaders concentrating in one region, and validators declaring fake locations. The study recommends σ ~= 10%, the highest value that keeps Asia's bins local, and γ between 3 and 7 (about 29 ms mean gain over random at γ = 7). These conclusions depend on today's geography.

## Micro data center company rolls out stackable data center for edge and AI

DevFeed: [Micro data center company rolls out stackable data center for edge and AI](<https://devfeed.tech/articles/micro-data-center-company-rolls-out-stackable-data-center-for-edge-and-ai-65236.md>)

Original publisher: [Read original article](<https://www.networkworld.com/article/4230865/micro-data-center-company-rolls-out-stackable-data-center-for-edge-and-ai.html>)

Author: Andy Patrizio

Published: 2026-10-05T19:27:41Z

Content type: news

Language: en

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

Topics: [data centers](<https://devfeed.tech/topics/data-centers.md>), [quasi-random network topology](<https://devfeed.tech/topics/quasi-random-network-topology.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [alerting](<https://devfeed.tech/tags/alerting.md>), [api](<https://devfeed.tech/tags/api.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [artificial-intelligence-data-center-data-center-design](<https://devfeed.tech/tags/artificial-intelligence-data-center-data-center-design.md>), [automation](<https://devfeed.tech/tags/automation.md>), [compute](<https://devfeed.tech/tags/compute.md>), [computing](<https://devfeed.tech/tags/computing.md>), [computing-capacity](<https://devfeed.tech/tags/computing-capacity.md>), [cooling](<https://devfeed.tech/tags/cooling.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [data-center-design](<https://devfeed.tech/tags/data-center-design.md>), [edge](<https://devfeed.tech/tags/edge.md>), [email](<https://devfeed.tech/tags/email.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [liquid](<https://devfeed.tech/tags/liquid.md>), [liquid-cooling](<https://devfeed.tech/tags/liquid-cooling.md>)

### AI overview

Zella DC introduced Zella Stack, a modular micro data center designed for edge deployments where space, installation time, or physical access can be challenging. Customers can add modules as capacity needs grow and configure air or liquid cooling, with options rated up to 124 kW. The system includes monitoring and automation through Zella Sense, which supports SNMP, Modbus, an API, email alerts, and several authentication methods.

### Source excerpt

Australian micro data center specialist Zella DC has introduced a modular data center platform designed to bring more computing capacity into locations where conventional racks and purpose-built facilities are difficult to deploy. Called Zella Stack, the new system uses a stackable architecture that lets customers begin with a relatively small amount of infrastructure and add capacity as computing requirements grow. The company unveiled the platform at Yotta 2026 in Las Vegas. The approach is aimed at edge computing deployments, where space, installation time and physical access can be significant challenges. The Zella Stack modules are 12RU containers that measures approximately 51.2 x 27.6 x 22.8 inches. The stack is made up of up to four modules for 48RU of hardware capacity. A wheeled base module contains the infrastructure needed to operate the system, while additional IT modules are stacked above it to provide space for compute equipment. The modular design also lets customers configure the system around different power and cooling requirements. Customers can use either air or liquid cooling depending on heat density. The company says cooling options can reach as much as 124kW, making the platform potentially suitable for substantially higher-density workloads than traditional small edge installations. At the top end of the product line is the Zella Stack 48, which consists of a base module and three IT blocks. The system weighs approximately 185 kilograms/408 pounds and stands about 2.48 meters tall (just over eight feet) when fully assembled. The operating noise level of approximately 50 decibels. The company says the individual components can be deployed rapidly, potentially allowing an edge data center to be operational within days rather than requiring the construction and commissioning process associated with a conventional data center. Zella's broader product portfolio is built around the same "data center in a box" concept. The company, formerly known