# Backblaze B2 Published articles for Backblaze B2. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Scaling AI Data Storage and Delivery with NVIDIA AIStore and B2 DevFeed: [Scaling AI Data Storage and Delivery with NVIDIA AIStore and B2]() Original publisher: [Read original article]() Author: Jeronimo De Leon Published: 2026-11-03T17:29:46Z Content type: article Language: en Sources: [Backblaze]() Topics: [AI Infrastructure](), [Cache](), [Amazon S3](), [big-data](), [data lake](), [Persistence](), [Isaac](), [sharding]() Tags: [ai](), [ai-data](), [ai-ml](), [ai-workloads](), [b2cloud](), [backblaze](), [backblaze-b2](), [capacity](), [cloud](), [cloud-storage](), [compute](), [curation](), [data](), [data-storage](), [datasets](), [delivery](), [fast](), [featured](), [featured-cloud-storage](), [local](), [nvidia](), [scaling](), [scaling-ai](), [storage]() ### AI overview Backblaze B2 holds the durable dataset record while NVIDIA AIStore caches and stages active data on storage near compute, balancing retention cost with faster access. For S3-based processing such as NeMo Curator, applications can use AIStore through configuration, prefetch data, and reuse cached reads. For Isaac GR00T training, AIStore stages a dataset revision to GPU-node disk and checks it against a manifest, leaving the trainer's data loader unchanged; checkpoints publish back through AIStore to B2. The post recommends sizing AIStore for concurrent working sets and reuse, using prefetch when predictable startup matters, keeping dataset revisions in write-once prefixes, publishing checksummed manifests, and measuring loading, warm reads, processing, and publication separately. The approach suits repeated or shared reads; a single sequential pass may be simpler streamed directly from B2. AIStore is infrastructure teams operate, and write-through adds local-write overhead before upload. ### Source excerpt AI workloads need durable data retention and fast access to active datasets. See how Backblaze B2 and NVIDIA AIStore pair a persistent capacity tier with local storage near compute for video curation, robot training, and repeated reads across shared datasets. The post Scaling AI Data Storage and Delivery with NVIDIA AIStore and B2 appeared first on Backblaze Blog | Cloud Storage & Cloud Backup ## System Design in 6 Points, Explained with a Chai Stall That Went Viral DevFeed: [System Design in 6 Points, Explained with a Chai Stall That Went Viral]() Original publisher: [Read original article]() Author: Shrey Vijayvargiya Published: 2026-10-28T09:00:00Z Content type: tutorial Language: en Sources: [iHateReading]() Topics: [system-design-project](), [Cache](), [nginx](), [least connections](), [Network]() Tags: [apache-2-0](), [api](), [asynchronous](), [aws](), [azure](), [backblaze-b2](), [backend](), [backups](), [batching](), [blockchain](), [bsd](), [cache](), [cache-control](), [cache-vs-queue](), [caching](), [cdn](), [coding](), [development](), [frontend](), [javascript](), [load-balancer](), [load-balancer-explained](), [logs](), [node-js](), [object-storage](), [product](), [programming](), [queue](), [react](), [read-replicas](), [read-replicas-explained](), [retries](), [retries-with-exponential-backoff-and-jitter](), [system-design](), [system-design-basics](), [system-design-cheat-sheet](), [system-design-for-beginners](), [typescript](), [web-development]() ### AI overview The article explains six system design patterns through a viral chai stall: cache repeated reads, queue slow background work, balance traffic across healthy app servers, use read replicas for read scaling, store files in object storage behind a CDN, and add timeouts, bounded jittered retries, and structured logs around network calls. It describes key limits: caches can serve stale data and need shared storage across servers; queues defer work and require durable jobs, idempotency, and monitoring; load balancers need shared sessions and can expose database connection limits; asynchronous replicas lag and do not reduce write load; file storage has provider-specific egress costs and private files need signed URLs; retries can duplicate unsafe operations or multiply across layers. Suggested order: logs and timeouts, object storage/CDN, then cache, queue, load balancer, and replicas after query and index fixes. ### Source excerpt System design basics in six points: cache, queue, load balancer, read replicas, object storage + CDN, and retries with timeouts and logs. One chai stall story,... ## Google Colab Pro vs. Runpod: Best GPU Cloud for AI Workloads DevFeed: [Google Colab Pro vs. Runpod: Best GPU Cloud for AI Workloads]() Original publisher: [Read original article]() Author: Zhen Lu Published: 2026-10-20T18:19:28Z Content type: comparison Language: en Sources: [Runpod Blog.]() Topics: [GPU](), [Google Cloud Platform (GCP)](), [Cloud Architecture]() Tags: [ai](), [ai-workloads](), [availability](), [backblaze-b2](), [benchmarking](), [billing](), [browser](), [cloud](), [cloud-costs](), [compare](), [comparison](), [compute](), [control](), [developers](), [google](), [gpu](), [gpu-cloud](), [pricing](), [reliability](), [vs]() ### AI overview Colab Pro and Pro+ sell metered compute-unit access, not a reserved GPU: hardware availability can change, premium GPUs are subject to availability, and different cards consume units at different rates. Colab runs generally last up to 12 hours; Pro+ supports continuous execution up to 24 hours with sufficient units, with idle timeouts and possible termination when units run out. Runpod lets users select a GPU and bills pods by the second while they run, making the hardware and duration more predictable; the excerpt distinguishes lower-cost Community Cloud from pricier Secure Cloud in vetted data centers. Colab offers convenient Google Drive integration. Runpod Cloud Sync supports several cloud storage providers but targets Google Cloud Storage rather than Drive; network volumes may suit data kept between pods, and the excerpt advises keeping irreplaceable data elsewhere. It suggests Colab for learning and prototyping, and Runpod when a specific GPU or a run that must finish matters. The supplied text is an excerpt, and its current-rate placeholder is unresolved. ### Source excerpt Compare Google Colab Pro and Runpod across pricing, reliability, and GPU access. Which is the better deal for developers running real AI workloads? ## I vibe coded my own Scrapingbee alternative and replaced my $249/mo plan with a $1/mo scraper DevFeed: [I vibe coded my own Scrapingbee alternative and replaced my $249/mo plan with a $1/mo scraper]() Original publisher: [Read original article]() Author: Pieter Levels Published: 2026-09-09T00:00:00Z Content type: opinion Language: en Sources: [@levelsio (Pieter Levels)]() Topics: [hosting](), [remote-development](), [X (Twitter)](), [alt:V](), [PaLM]() Tags: [ai](), [automation](), [backblaze](), [backblaze-b2](), [backups](), [cloudflare](), [domains](), [email](), [hetzner](), [llms](), [moderation](), [open-source](), [playwright](), [r2](), [saas](), [scraper](), [scraping](), [scrapingbee](), [storage](), [stuff-i-made](), [vibe-coding](), [vps](), [xai]() ### AI overview The author reports replacing a $249/month Scrapingbee plan with a self-built scraper costing about $1/month on a VPS. OpenSERP was blocked on the author's VPS and laptop; the author also says residential proxies from IPRoyal were frequently blocked and vanilla headless Chrome was detected. A Playwright setup using real Google Chrome in headful mode, a persistent context, slow pacing, and different residential IPs reportedly reached about 90% success, similar to or better than Scrapingbee. These are the author's reported results; the IP recommendations are not disclosed. ### Source excerpt ✨ Okay with lots of help from @javilopen and his Spanish scraping friends I've managed to vibe code my own @Scrapingbee alternative and replace my $249/mo Scra... ## Incremental backup on AWS Glacier with rustic DevFeed: [Incremental backup on AWS Glacier with rustic]() Original publisher: [Read original article]() Published: 2026-08-19T21:04:01Z Content type: tutorial Language: en Sources: [Personal blog of Anurag Bhatia]() Topics: [backups](), [restic](), [Amazon Web Services](), [Amazon S3](), [Rust](), [Open Source](), [Feathers]() Tags: [aws](), [aws-glacier-deep-archive](), [backblaze-b2](), [backup](), [backups](), [client](), [config](), [cost](), [data](), [encryption](), [encryption-at-rest](), [incremental](), [lifecycle](), [open-source](), [password](), [restic](), [restore](), [rust](), [rustic](), [s3](), [snapshots]() ### AI overview A practical guide to using rustic, a Rust implementation compatible with restic, for incremental backups with AWS Glacier cold storage. It explains keeping backup data in cold storage while maintaining metadata in S3 Standard or Backblaze B2 so snapshots, incremental backups, and targeted restores remain possible. ### Source excerpt I have been a heavy restic user for several years. In Jan of this year, I tried making restic work with AWS Glacier deep archive but it did not work out. AWS Glacier Deep Archive, for those who may not know, is one of the cheapest storage options, with costs as low as $1/TB, but with very expensive retrieval due to high egress charges from AWS, in addition to retrieval fees, API charges, etc. It can act as a backup of last resort, kind of like insurance if everything else fails, since it costs roughly 1/7th of what Backblaze B2 charges for the same data ($6.95/TB/month). Restic by design is not compatible with Glacier because it stores a mix of data blobs and tree blobs inside /data. Thus, as AWS lifecycle rules move data from S3 Standard to Glacier, basic repo operations start failing. They do have some experimental support as per FAQ, but it did not work out for me when I tried in Jan. A few days back, I came across an open-source project named rustic -- it's a restic implementation in Rust and supports cold storage. The rustic client is compatible with a traditional restic repo. Rustic and cold storage Rustic supports AWS Glacier as well as OVH cold storage. The way it works is: they keep all data in cold storage and a copy of the metadata in hot storage. This additional copy of metadata in hot storage allows operations like incremental backups, viewing snapshots, and requesting restores where only the specific data blobs needed for the restore are pulled from Glacier. This additional hot storage can sit on S3 standard or even Backblaze B2 (which I prefer due to zero API cost). Demo 1) Create repo Let's create two buckets - one in S3 (to store data in Glacier) and one on Backblaze B2 to hold an additional copy of metadata. > rclone mkdir s3:rustic-demo-cold > rclone mkdir b2:rustic-demo-hot 2) Create rustic config profile This file will define the cold and hot repo locations, along with the password file location, restoration command syntax, etc. [repository] repo ## The Subtle Magic of tsnet DevFeed: [The Subtle Magic of tsnet]() Original publisher: [Read original article]() Author: Xe Iaso Published: 2023-07-07T00:00:00Z Content type: article Language: en Sources: [Tailscale Community]() Topics: [tailscale](), [Network](), [Cloud Networking](), [software-development](), [cdnjs](), [ai-image-generator]() Tags: [api](), [authentication](), [authorization](), [automatic1111](), [backblaze](), [backblaze-b2](), [backend](), [boltdb](), [bot](), [c](), [c-sharp](), [cache](), [caching](), [cdn](), [certificates](), [chatgpt](), [services](), [tailscale](), [tailscale-up-2023](), [talk](), [tsnet]() ### AI overview tsnet embeds Tailscale in Go services, giving them a tailnet IP address, DNS name, HTTPS certificates, and access controls through ACL tags. Xe's examples show how this can simplify internal service discovery and access: XeDN uses tailnet access for its control API and metrics; Robocadey connects a Fly.io bot to a homelab Stable Diffusion server through a tsnet HTTP client, avoiding public exposure of the homelab; and tclip uses Tailscale identity for paste attribution and Funnel to share selected content publicly while keeping submission and paste listings private. The talk also describes proxy-to-grafana and libtailscale, a C wrapper intended to let non-Go programs embed tsnet; at the time of the talk, libtailscale was early-stage, with Python and Ruby tested most. The examples reflect the author's setup and date from May 2023; the repost notes product and code information may be out of date. ### Source excerpt Xe's Tailscale Up talk about how to use tsnet in your services, complete with many personal examples.