Product
Arcois Exera
An S3-compatible gateway that keeps every piece of S3 metadata itself and leaves your stores holding bytes. Everything on this page ships today; the one exception carries a badge.
Aggregation into sealed volumes
Small objects are appended to an open volume on the pod's local disk and sealed at a size or age you set. Each volume carries its own manifest, so its contents can be read back from the volume alone. Reads fetch one byte range. Large objects skip packing and go straight through.
Deleted objects free capacity when their volume is compacted, so reclamation is background work that follows the delete.
Five typed, checkpointed operations
Import an existing bucket lazily (metadata first, bytes on first read), convert its layout, drain a store, rebalance after adding one, or cut over from one provider to another with near-zero downtime. Every destructive checkpoint waits for your approval.
Placement across stores, from any vendor
A placement policy names groups of S3-compatible stores. Across groups is replication, within a group is sharding. Backend health, first-group acknowledgement, an owed-copy index and a healer keep copies level. Lifecycle Transition rules move objects between placements by storage class.
Between two sites, active-passive geo-replication ships every change through an in-transaction outbox, with epoch fencing, an optional witness, and failover you trigger.
Realms, quotas, metering, IAM
A realm isolates one tenant's buckets, users, credentials and policies. Quotas warn softly then refuse. Daily usage snapshots feed billing. The policy engine is default-deny with identity and bucket policies, public-access block and conditions. Tenants use the standard IAM and STS tooling on the S3 port.
Tape archive
Arcois ships an LTFS tape service for TS4300 and TS4500 libraries: inventory, formatting, verification, two-copy archive and retrieve, over an HTTP and gRPC API. For Exera customers it is enabled per deployment: sealed volumes are sent to tape on a schedule, so drives stream. Details.
The API your clients already speak
Objects, buckets, multipart, versioning, listing V1 and V2, conditional headers and conditional writes, tagging, lifecycle, access logging, event notifications, presigned URLs, virtual-host addressing, five additional checksums, and three server-side encryption modes. Tested continuously against the public S3 test suites and the common SDKs and tools.
Security and resilience
Built for a security review, not just a demo
Every request verified
Full Signature Version 4 including chunked uploads. Anonymous access is policy, never a bypass. Legacy ACLs are refused rather than half-honoured.
Encryption and retention
SSE-S3, SSE-C and SSE-KMS with a pluggable key provider. Object Lock in governance and compliance modes, legal hold, and versioning.
Secrets and configuration at rest
A passphrase-rooted cluster key encrypts stored credentials and seals a periodic configuration backup to a system bucket. A lost control plane is restored from one surviving store and the passphrase.
Your certificate authority
Outbound trust to stores, identity providers and peers is configured with your own CA. Private CAs are first-class, not a flag you hope exists.
Nothing phones home
Objects stay on your stores, metadata in your cluster. Every outbound connection goes to a destination you configured. Arcois receives no telemetry.
Durability as a contract
A torn write is never read as complete. No read location is ever dropped. Owed copies are tracked, not forgotten. The limits are written down in the documentation, not discovered.
Operating it
Console, CLI, metrics and written runbooks
Operator console
Fourteen screens: overview, usage and metering, realms, stores, placements, buckets and objects, credentials, operations, disaster recovery, geo-replication, IAM, administrators and identity providers, events, maintenance.
Operator CLI
Every console action has a command. Runbooks are written CLI-first, so what you rehearsed is what you run at three in the morning.
Metrics and diagnostics
Prometheus metrics across requests, volumes, stores, metadata plane, migrations and per-realm usage. Structured logs with request IDs and a diagnostic bundle endpoint.
At a glance
What you can hold us to
| Capability | Status | Note |
|---|---|---|
| Aggregation and object-count relief | ships | Core of the product, heavily tested |
| Migration between S3-compatible providers | ships | One DNS switch; the old copy stays until you remove it |
| Redundancy across S3-compatible stores from different vendors | ships | Health, owed-copy index, healer |
| Geo-replication between two sites | ships | Active-passive, failover you trigger |
| Multi-tenant S3 as a service | ships | Realms, quotas, usage snapshots, IAM and STS |
| Storage-class tiering between placements | ships | Lifecycle Transition rules |
| Tape archive (Arcois tape service) | integrated on request | Scheduled volume archiving per deployment |
Exera works with any S3-compatible object store, from any vendor, one or several at a time.
See it on your object sizes.
A demo starts from your distribution, not ours.