Storage
Benchmark
Reduce Storage by 43.9% with a Custom Zstd Dictionary
Train on real GitHub events and measure storage savings, dictionary overhead, and S3 read performance.
SeaweedFS Enterprise
One cluster.
Objects, tables,
vectors, and files.
Fast distributed storage, with the protection and control your production data needs.
Serve S3 objects, Iceberg tables, vector search, and POSIX workloads from the same engine. Add recovery, storage efficiency, and operational tools to the open-source core.
Free for development and testing under 25TB.
One storage platform
01S3Objects
02IcebergTables
03VectorsEmbeddings
04POSIXFiles
SeaweedFSOne distributed storage engine
MetadataNamespace & catalog
Volume serversDistributed object data
RecoveryErasure codingTiering
Your infrastructure. Your data.
Trusted by
Cisco
NVIDIA
Proton
TrueNAS
ÆnixOne platform
Built for your workloads.
01 / OBJECTS
S3 object storage
Use familiar S3 clients and tools with a distributed store that grows with your data.
Start a cluster 02 / ANALYTICSIceberg table storage
Query table buckets through a built-in REST catalog, with maintenance in the background.
Connect your engine 03 / AI & SEARCHS3 vector buckets
Store embeddings, filter by metadata, and run similarity searches through the S3 Vectors API.
Explore vector buckets 04 / FILESNative file access
Mount SeaweedFS as a Linux filesystem and bring existing file-based applications.
Meet the kernel mountThe Enterprise layer
Protect your data.
Make your hardware go further.
01
Recover with confidence
Undo deletes and overwrites, and repair damaged data when hardware fails.
02
Store more, use less
Reduce data and metadata overhead, then reclaim space when it is no longer needed.
03
Keep operations visible
See how the cluster is performing and manage the people and workloads using it.
Measured, with the details
See what changes in practice.
43.9%less payload storage
Custom zstd dictionaries versus plain zstd on 1,000 GitHub events. Dictionary overhead excluded.
Read the benchmark 18–58×less filer metadataSealed directories on measured live-cluster workloads. Savings depend on directory layout.
See the measurements ~5×faster S3 readsRDMA versus HTTP for a 64 MiB GET in the published lab benchmark. Results depend on the workload.
Explore S3 over RDMASimple capacity licensing
Scale on your own terms.
Self-hosted storage. A per-TB software license.
Choose support separately to match your team.
$2/ TB / month
or $20 / TB / year
Hardware and infrastructure provided by you.From the engineering team
Learn by building.
AI & ML
Benchmark
S3 over RDMA Is 5× Faster Than HTTP
Compare S3 reads over RDMA and HTTP with measured throughput and byte-for-byte verification.
Analytics
Walkthrough
One Table Bucket, Many Engines: DuckDB and Trino via databow
Run the same SQL against one table bucket using Trino and DuckDB, with measured timings.
Get started
Put your data to work.
Evaluate Enterprise free for development and testing under 25TB.