Comparison of ZeroFS and Amazon S3 Files Reveals Key Differences

Overview of ZeroFS and Amazon S3 Files

ZeroFS and Amazon S3 Files both expose POSIX compliant file systems backed by object storage but have contrasting internal layouts. The choice of implementation affects how files are handled, stored, and accessed. S3 Files aims to keep the identity of ordinary S3 objects intact, whereas ZeroFS allows for enhanced internal persistence and storage efficiency.

Storage Layout Differences

S3 Files utilizes Amazon EFS and keeps a one-to-one mapping between files and S3 objects. When a file is updated, changes reflect back to the corresponding S3 object. Active data resides in a low-latency tier classified as high-performance storage by AWS. In contrast, ZeroFS uses a more complex structure where file contents are split, compressed, and encrypted, leading to a loss of direct file visibility through the S3 API.

Interoperability and File Access

S3 Files ensures that files written through it become standard S3 objects after a designated export period, facilitating easy access and visibility with standard S3 tools. Conversely, ZeroFS lacks this feature, meaning files modified in ZeroFS are not directly accessible via the S3 API, making it suitable for specific internal applications that do not require immediate S3 visibility.

Cold Access Procedures

When accessing S3 Files, the first request can trigger an import process where metadata for every object is loaded, with files below a certain size being copied into the high-performance tier. AWS notes that initial listings involving a large number of objects may experience delays, which could affect performance in certain usages.