Tigris Data
An S3-compatible, globally distributed object storage platform engineered to optimize small-object workloads through metadata inlining, adjacent key coalescing, and LSM-backed caching, delivering sub-10ms read/write latency for KB-sized payloads.
Summary
An S3-compatible, globally distributed object storage platform engineered to optimize small-object workloads through metadata inlining, adjacent key coalescing, and LSM-backed caching, delivering sub-10ms read/write latency for KB-sized payloads.
An alternative S3-compatible provider positioned between hyperscaler general-purpose storage and specialized caching layers. Addresses the latency and API cost penalty that standard S3 imposes on workloads dominated by millions of small objects such as log aggregations, ML feature stores, and IoT telemetry.
- Not a CDN or cache — it is a durable object store with full S3 API compatibility.
- Small-object optimization does not mean it sacrifices large-object throughput; the architecture handles both.
- Global distribution does not imply eventual consistency for all operations — write consistency is maintained per-region.
implementsS3 API — full S3 compatibility for existing toolingsolvesSmall Files Problem — metadata inlining bypasses per-object storage overheadsolvesRequest Amplification — key coalescing reduces API call volume for adjacent objects
Definition
An S3-compatible, globally distributed object storage platform engineered for small-object workloads. Uses metadata inlining, adjacent key coalescing, and LSM-backed caching to deliver sub-10ms read/write latency for kilobyte-sized payloads.
Standard S3 imposes severe latency and API cost penalties on workloads dominated by millions of small objects (logs, ML feature stores, IoT telemetry). Every object incurs per-request overhead regardless of size. Tigris inlines small payloads directly into the metadata database layer, bypassing the per-object storage overhead entirely.
Low-latency storage for ML feature stores, log aggregation pipelines, IoT telemetry ingestion, and any workload where object sizes are consistently below 1MB.
Connections 4
Outbound 4
Resources 2
Reproducible YCSB benchmarks proving sub-10ms latency for small objects versus AWS S3 and Cloudflare R2, quantifying the metadata inlining advantage.
Official documentation for the globally distributed S3-compatible object storage platform, covering architecture, API compatibility, and deployment patterns.