今日どうしても使ってみたくてS3filesをEC2にマウントしてみたので、あとでブログにしてみます！コンテナに対しての永続化とかで使われたりするのかな？あんまり実用性の部分がわからなかった🤣 あとは複数AZに対してのマウントが前提で使われそうな感じもしたな🤔 今度色々やってみます！！

I just published The 250MB Wall: Why We Trashed Lambda Layers for a Better Way to Ship Binaries medium.com/p/the-250mb-wall-… #aws #s3Files #lambda

【Tech Blog 公開✨】 「S3をマウントして使いたい」 それはAWSエンジニアが夢見つつ、パフォーマンスや制約の壁に幾度も跳ね返されてきた、永遠のテーマかもしれません。 e-dashプロダクト開発部の伊藤（a2-ito）さんが、2026年4月にリリースされた『S3 Files』の実力を徹底検証しました！ 並列アクセスで叩き出した「リード約33.5GB/s」という異次元の速度と、単価がS3標準の「約14倍」というシビアな現実。単なる「便利」で終わらせず、性能とコストの境界線を冷静に暴きます。 インフラの制約を嘆く時代は終わり、強すぎる武器をどう選び抜くか。エンジニアの真価は「構築」から「選定」へとシフトしています。 👇記事はこちら zenn.dev/edash_tech_blog/art… #AWS #S3Files #EC2 #S3

🆕 My latest blog on S3 Files! For nearly two decades, "S3 as a file system" has meant download to /tmp, process, upload, and a copy-then-delete dance to fake a rename. My new post looks at AWS's S3 File - the managed NFS layer that finally retires that pattern. I set up two equivalent Lambdas via Terraform: one written the boto3 way, and one that just calls open() and os.rename() against a mounted file system. It's the same workload, same bucket, but ~5x faster on the after side with materially less code to maintain. The post also covers when S3 Files is the wrong answer. Mountpoint still wins for large sequential reads at 459 MiB/s, and there's an access point ownership issue I ran into plus a 60-second commit delay you have to design around. lckhd.eu/sWVVlW #aws #lambda #s3files #serverless

#AWS introduced #S3Files, letting users mount an Amazon S3 bucket and access data through a standard file system interface. Apps can read & write files with standard file operations, while the system translates them into S3 requests. Find out more: bit.ly/4cwWOpK

📰💡 AWS S3 is now a high-performance file system. Introducing S3 Files! Mount S3 buckets natively to EC2, ECS, or EKS via NFS without refactoring. ⚡ No more data copying 🚀 ~1ms latency (EFS tech) 🔒 Native IAM POSIX controlbs A massive win for ML pipelines! #CloudBreach #AWS #CloudNative #S3Files #CloudSecurity #DevSecOps #s3bucket

We shipped S3 Files into production 5 days ago for @cfosilvia. Mount a skills bucket at /mnt/s3files on ECS, expose it through a bash API, let agents search docs with standard POSIX tools. No SDK. No object downloads. Just grep on a mount point. It really is goated.

For a hot minute I was getting a little jelly over the new AWS S3Files product. Its certainly in the "magic quadrant" (for Amazon). Basic Memory Cloud uses @flydotio and @TigrisData for hosting and bucket storage. They are much more affordable and so much easier to use.

linkedin.com/pulse/amazon-s3… on my view, #s3files vs. #juicefs

1/ S3 Files just launched. Your Lambda functions can mount S3 buckets as a local file system. open(), pathlib, os.walk(). No boto3 for file access. I built AI agents on it. Here's what I learned 🧵 2/ The example: point it at a public GitHub repo. A durable function clones the repo to the S3 Files mount. Two Strands agents (security style) review the code in parallel. All three Lambda functions share the same mounted bucket. 3/ The IaC was the real adventure. S3 Files is brand new. The CloudFormation types (AWS::S3Files::FileSystem, MountTarget, AccessPoint) aren't in cfn-lint yet. The trust policy uses elasticfilesystem.amazonaws.…. The access point needs CreationPermissions, not CreationInfo. 4/ The full SAM template deploys everything: VPC, S3 Files, mount targets, access point, durable orchestrator, two Strands agents, async API Gateway. One sam deploy and you're running.

I swear people REALLY underestimate the filesystem aspect of this. Having your memory be portable across models is critical and the unix commands available in the CLI are so ergonomic to the model. This is what has me so excited about the S3Files launch

Didn’t even mention pricing, S3Files is insanely expensive

Amazon #S3Files went GA on April 7th, and while there are plenty of setup guides floating around, the article below takes a more architectural lens. It digs into what layers in your stack become unnecessary when S3 buckets can be mounted as NFS file systems from EC2, Lambda, EKS, and ECS. There is a good discussion included on the tradeoffs involved. The ~60 second commit interval, rename costs, the 50 million object threshold, VPC requirements. It maps out where S3 Files genuinely removes friction and where you still need to design carefully around its constraints. Check out this really good article from Kento Ikeda! #s3 #aws lckhd.eu/rnoimX

S3Files、EFSからコスト観点で置き換えるくらいしか考えてなかったけど、そもそもS3上のファイルをコンピューティングリソースから参照したいケースがあり、それがファイルシステムのように扱えるようになることで色々ハッピーと理解しました。

☁️🗂️ Amazon S3 wird „dateifähig“: S3 Files ist da Objektspeicher in S3 – aber dateibasierte Tools erwarten ein Dateisystem? Genau dieses Dilemma soll S3 Files lösen: Amazon macht S3-Buckets als vollwertiges Dateisystem nutzbar – ohne Copy/Synch-Pipelines und ohne Daten zu duplizieren. Was steckt dahinter?  ✅ Zugriff auf S3-Daten wie auf ein Filesystem (Mount)  ✅ Dateisystem-Operationen werden automatisch in S3-API-Calls übersetzt  ✅ Parallelzugriff für viele Computer-Instanzen (EC2/Container/Serverless)  ✅ Besonders spannend für Data Engineering und KI/ML 👉 it-administrator.de/amazon-s… #AWS #AmazonS3 #S3Files #CloudStorage #CloudComputing #DataEngineering #MLOps #AI #DevOps #Sysadmin #ITAdministrator

A really big S3 announcement - #S3 Files!! If you've ever had to copy data out of S3 just so a file-based tool could use it, you know the frustration. S3 Files lets you mount S3 buckets as file systems directly, no data duplication, no sync pipelines, no workarounds. Your data stays in S3 while apps access it natively. This looks really useful and a nice improvement over previous approaches like S3 Mount Point. AI agents, ML pipelines, and traditional file-based applications can all read and write to S3 as a file system without code changes. #aws #S3Files lckhd.eu/yt6Tz5

Do this for every database you use. Hard delete is not good for non-personal projects. Created, updated and deleted date columns are good to have when data is processed outside main database, for example datawarehouse or just spark on s3files. Hard delete = must use CDC, hard.