Local lake workflows without extra services.
Use S3-compatible storage and an Iceberg REST Catalog for DuckDB-oriented development, internal data platform sandboxes, and lakehouse integration testing.
Apache 2.0 storage infrastructure
One binary for object, file, block, and Iceberg storage.
GrainFS is a single-binary distributed storage server for developer and data infrastructure teams that need S3-compatible object storage, NFS/9P file access, Linux NBD volumes, and an Iceberg REST Catalog over one storage layer.
grainfs quick start
# start a local S3-compatible storage server
DATA_DIR=./tmp
make bin/grainfs
CLUSTER_KEY=$(openssl rand -hex 32)
./bin/grainfs serve --data "$DATA_DIR" --port 9000 --cluster-key "$CLUSTER_KEY"
# use it from standard S3 tooling
echo "hello grainfs" > file.txt
aws --no-sign-request --endpoint-url http://localhost:9000 s3 cp file.txt s3://default/Four storage interfaces. One operational unit.
GrainFS exposes object, file, block, and table/catalog surfaces over shared storage and cluster contracts. Protocol adapters translate client behavior without turning every interface into a separate system to run.
S3-compatible HTTP APIobject
NFSv4 + 9P2000.Lfile
single binary
GrainFS
Linux NBD volumesblock
Iceberg REST Cataloglake
Built for teams stitching storage systems together.
GrainFS starts with open-source adoption, then grows into paid operational tooling for teams that need confidence around deployment, recovery, observability, and upgrades.
Small clusters with multiple protocol surfaces.
Run object, file, and block interfaces near private infrastructure, GPU boxes, lab devices, or edge deployments where a heavyweight storage stack is too much.
A storage protocol testbed in one binary.
Replace a pile of local fixtures with one storage server for S3, NFS/9P, NBD, and Iceberg client and product integration tests.
Operator trust is the product boundary.
The roadmap focuses on safe defaults, explicit escape hatches, observable recovery, compatibility gates, and operational runbooks — the parts teams need before trusting storage with real workloads.
548.30MiB/s
Same-host single-node warp run, 64 KiB objects, concurrency 32, signed requests, 0 errors.
1849.34MiB/s
Warm-read pass over objects kept from the preceding PUT pass on the local Apple M3 setup.
60k+ops/s
Single-node Iceberg REST Catalog benchmark via warp iceberg, with 0 errors in the referenced run.
Apache2.0
Permissive open-source core for local use, small clusters, protocol testing, and infrastructure adoption.
Benchmark claims should be read with their methodology: same host class, comparable durability, object size, concurrency, warm/cold-cache rules, signed S3 requests, and raw artifacts. See the benchmark reference in the GrainFS repository before publishing comparisons.
Open core adoption. Paid operational confidence.
The business model starts with a strong open-source engine, then monetizes the control plane, enterprise operating features, and support paths that production teams need.
Open-source engine
- Single-node and small-cluster usage
- S3, NFS/9P, NBD, and Iceberg surfaces
- CLI, docs, runbooks, and benchmark references
Team control plane
- Cluster dashboard and health inventory
- Diagnostics, backup/restore workflows, upgrade checks
- Slack/email support and operational guidance
Private infrastructure support
- Air-gapped install and long-term support builds
- SSO/RBAC, audit workflows, recovery assistance
- Architecture review for on-prem, edge, and lab clusters
grainfs.gritive.com
Try GrainFS locally in minutes.
Start with the open-source binary, validate your storage workflow, then bring the control-plane and enterprise operating model when the workload matters.