S3 Files becomes POSIX mount

Most cloud storage makes you pick a side. You keep data in Amazon Simple Storage Service because it is cheap and durable, or you move it into a file system because old code expects folders, locks, and `open` calls. (aws.amazon.com) A file system is the thing behind paths like `/data/images/cat.jpg`. It gives programs concrete rules for reading, writing, renaming, and locking files, which is why Linux tools and a lot of machine learning code still assume one exists. (docs.aws.amazon.com) Object storage works differently. Amazon Simple Storage Service stores blobs behind bucket names and object keys, which scales extremely well, but it does not natively behave like the shared disk that many applications were written for. (aws.amazon.com) Amazon Web Services spent years offering a partial bridge called Mountpoint for Amazon Simple Storage Service. Mountpoint let Linux machines mount buckets, but Amazon’s own documentation said it did not implement the full Portable Operating System Interface standard and was best for high-throughput jobs that could live without full file-system features. (docs.aws.amazon.com) Now Amazon Web Services has launched Amazon Simple Storage Service Files. The service makes a bucket accessible as a shared file system with full file-system semantics and about 1 millisecond latency, while keeping the data in Amazon Simple Storage Service instead of copying it into a separate store first. (aws.amazon.com) Portable Operating System Interface compatibility is the key phrase here. It means software can do ordinary file operations like open, read, write, rename, and lock against data in a bucket, instead of being rewritten around object-store application programming interfaces. (docs.aws.amazon.com) Amazon says the service is built using Amazon Elastic File System. In practice, that means Amazon Simple Storage Service Files uses mount targets and access points like a network file system, while Amazon Simple Storage Service remains the system of record underneath. (docs.aws.amazon.com) The performance trick is that only the active working set sits on high-performance storage. Amazon says the service can deliver low latency and multiple terabytes per second of aggregate read throughput by caching the files teams are actually touching, instead of preloading an entire bucket. (aws.amazon.com) That changes a very specific pain point in data pipelines. Teams often stage training data, logs, or media assets out of Amazon Simple Storage Service into a separate file system before a job starts, then sync results back later, which adds copies, delays, and another thing to break. (aws.amazon.com) Amazon is aiming this at compute that already lives inside its cloud. The launch materials say Amazon Simple Storage Service Files can be shared across Amazon Elastic Compute Cloud, Amazon Elastic Kubernetes Service, Amazon Elastic Container Service, Amazon Lambda, Amazon Fargate, and Amazon Batch, with support for up to 25,000 active connections. (aws.amazon.com) The Lambda detail is unusually concrete. Amazon’s setup guide says a Lambda function must sit in the same virtual private cloud as the mount target, and the console can automatically create one mount target in every Availability Zone in the default virtual private cloud plus an access point for the file system. (docs.aws.amazon.com) The bigger shift is conceptual. Amazon Simple Storage Service used to be the place you parked objects, and Amazon Elastic File System was the place you mounted directories; after April 7, 2026, Amazon is selling a hybrid where the bucket is still the bucket, but the same data can also look like a normal shared disk to software that never learned object storage. (aws.amazon.com)