Amazon S3 vs XtreemFS

Amazon S3 is an object storage service for storing, managing, analyzing, and protecting data used by data lakes, cloud-native applications, and mobile apps. Capacity adjusts as objects are added or removed, without requiring storage to be provisioned in advance. Storage classes cover frequent access, infrequent access, and archiving. Versioning preserves object versions, while Replication can copy objects to buckets in the same or different AWS Regions. Amazon Athena and Redshift Spectrum can query S3 data in place with SQL; S3 Vectors adds vector storage and querying, including integration with Amazon Bedrock Knowledge Bases. Security controls include encryption for uploads, Block Public Access, and Object Lock for customer-defined retention periods. AWS documents access through its Management Console, CLI, SDKs, and REST API. S3 is listed for Android, API, iOS, Linux, macOS, web, and Windows. A Free Tier is available. Pricing is usage-based, with charges varying across storage, requests, retrieval, transfer, and other features. Some archived objects have minimum storage durations of 90 or 180 days, and early deletion incurs a prorated charge.

XtreemFS is a free, fault-tolerant distributed file system for general-purpose storage, available for Linux, macOS, Windows and self-hosted deployments. Files are replicated across storage servers, with automatic failover to a backup replica if the primary fails. Full or partial read-only replicas support immutable files and can be used to create caches across datacenters. The project says an installation can grow to thousands of storage and metadata servers by adding or removing storage servers. POSIX compatibility makes it usable as an NFS replacement. SSL and X.509 certificate support are built in, and the site says XtreemFS can run over the internet without a VPN. A Hadoop client supports use as an HDFS replacement. Listed uses include filer replacement, HPC and Hadoop clusters, VM block storage and cross-branch sharing. Linux distribution packages and precompiled Windows and Mac OS X clients are available. The FAQ says a single volume may hold up to tens of millions of files, with no known limits stated for volumes and services. The project is open source under a BSD license and is developed by Zuse Institute Berlin.