Storage and Recovery

Storage and recovery sit beneath Products and Studio as core platform services. Workloads request typed volumes, objects, artifacts, and backup operations while GlobalStacks manages content-addressed snapshots, retention, verification, and restore without distributing raw infrastructure credentials.

Sandbox volumes

Attach durable filesystem state through cluster-supported storage backends.

Object stores

Use S3-compatible services, including extension-provided RustFS capacity, through typed facilities.

Immutable artifacts

Address build and release payloads by purpose and content digest.

Backup and restore

Create encrypted, deduplicated snapshots and apply retention, restore, pruning, garbage collection, and verification as managed lifecycle operations.

How backup data moves

Store changed data, not another full copy.

GlobalStacks turns each backup into an ordered manifest of content-addressed chunks. Existing chunks are reused inside the same tenant and destination key boundary; only new content is encrypted and uploaded.

1. 01

   Chunk

   FastCDC content-defined chunking keeps boundaries stable when files change.

2. 02

   Identify

   Each chunk receives a SHA-256 digest used for lookup and integrity checks.

3. 03

   Protect

   New chunks are encrypted with AES-256-GCM before they reach object storage.

4. 04

   Commit

   An ordered manifest records the chunks, offsets, lengths, and snapshot identity.

Deduplication

Reuse without crossing trust boundaries.

Identical chunks are uploaded once and referenced by every snapshot that needs them. Deduplication is scoped to one tenant and destination encryption key, so storage savings never become a cross-customer data channel.

Logical bytes

The full size represented by retained snapshots.

Uploaded bytes

The new encrypted data transferred for each run.

Unique chunks

Content currently stored after deduplication.

Dedup ratio

Logical size compared with newly stored data.

Recovery lifecycle

Retention is separate from deletion.

Expiring a snapshot removes its references first. Garbage collection deletes an object only after no retained snapshot refers to it, preserving shared chunks for every recovery point that still depends on them.

Policy and scheduling

Bind backup destinations, retention windows, and manual or scheduled runs to typed resources.

Reference-safe pruning

Mark expired snapshots as pruned and decrement their chunk references without touching retained recovery points.

Garbage collection

Measure reclaimable chunks and bytes, then remove only unreferenced objects and catalog records.

Verified restore

Fetch the manifest in order, decrypt each chunk, verify its digest, and fail closed when data is missing or corrupt.

Clone recovery

Restore supported application configuration and volume metadata into a new target rather than overwriting the source implicitly.

Operational evidence

Record run status, checksums, logical and uploaded bytes, restore results, prune counts, garbage collection, usage, and audit events.

Bring or consume storage

One recovery model across storage providers.

Target a ready, purpose-scoped S3-compatible object store or GlobalStacks-managed backup capacity. Credentials stay behind broker references and operation-scoped access instead of being copied into workloads.

Explore the marketplace