Skip to content
RenderCache

Closed infrastructure for CG studios

Enterprise 3D Asset Synchronization

OpenVDB and Alembic simulation caches and 8K textures move between your remote render farms through dedicated cache nodes — encrypted, resumable from the exact point of failure, with no manual copying.

API documentation
Encryption
AES-256-GCM
Node link
up to 25 Gbit/s
Chunk
8 MiB · BLAKE3
Node Graph NH_0420 / sync_main
fluid_sim_v04 · chunk map 1,873 / 2,048
transferred resuming queued ETA ~0:58
Fits your pipeline HoudiniNukeMayaUnreal EngineKatanaBlenderDeadlineOpenCueOpenUSD

01 / Features

Infrastructure that understands caches

General-purpose file storage spends hours re-copying sparse volumes and per-frame sequences. RenderCache reads the formats your farms produce and sends only what changed.

AES-256

AES-256-GCM encryption

Every chunk is encrypted on the studio side before it leaves. Project keys stay inside your perimeter or HSM; cache nodes and our servers only ever see ciphertext — which satisfies studio NDAs and client security audits.

  • Key exchangeX25519
  • Key rotationper project
  • TransportTLS 1.3 · mTLS
Resume Broken Transfers

Resume broken transfers

Files are split into content-addressed chunks. If the link drops at 97% of a two-terabyte cache, the agent continues from the last verified chunk and pulls the missing ones from the nearest cache node.

  • Chunk checkBLAKE3
  • Upload sessionkept 7 days
  • Parallelismup to 64 streams
VDB Support

Native OpenVDB support

The agent reads the VDB tree and diffs leaf nodes between versions. Re-simulate frames 1100–1180 and only the changed voxels of those frames go over the wire — not the whole sequence.

  • Grid diffper leaf node
  • Sequences$F4 · #### · %04d
  • VersionsOpenVDB 9–12

.abc

Alembic by range

Only changed time samples and hierarchy objects are sent.

8K · UDIM · .tx

Tile-level textures

UDIM tiles and MIP levels are deduplicated across asset versions.

msk · spb · nsk

Regional cache nodes

A hot cache next to the farm: repeat reads never touch the studio uplink.

REST · Python SDK

Render manager hooks

Deadline, OpenCue and Afanasy trigger a sync as soon as a cache is written.

02 / Architecture

Studio → cache node → render farm

  1. STEP 01

    rcd agent at the studio

    Watches cache directories, splits files into chunks, encrypts them and publishes a version manifest.

  2. STEP 02

    Regional cache node

    Stores encrypted chunks next to your farms and serves every subscribed site. Upload once — every farm reads locally.

  3. STEP 03

    The farm assembles the cache

    The farm agent verifies each chunk's BLAKE3 hash, decrypts and rebuilds files at their original paths before the render task starts.

rcd — farm-spb-02bash
$ rcd pull NH_0420/fx/fluid_sim_v04 --frames 1001-1240
[10:42:07] manifest v14 · 2,048 chunks · 15.9 GiB
[10:42:07] source: edge-spb-1 (rtt 3 ms)
[10:46:51] link lost · 1,873/2,048 verified
[10:47:20] resume session us_8Hq2 from chunk 1,873
[10:48:18] done · 175 chunks · BLAKE3 ok · AES-256-GCM
$ ▍
SLA
99.95%
on the Enterprise plan
Regions
3 + on-prem
Moscow, St Petersburg, Novosibirsk
Traffic saved
up to 92%
on repeat VDB versions
Data residency
RU
account data under 152-FZ

03 / Pricing

Pay for cache capacity, not for copies

Your plan sets the hot-cache capacity and the number of connected agents. Traffic between a cache node and your farms inside the same region is free.

Team

$330/ mo

billed monthly

Small studios working with one external farm.

  • —10 TB hot cache
  • —Up to 12 agents
  • —1 region
  • —VDB, Alembic, 8K/UDIM
  • —Email support, next business day
Studio pick

Studio

$990/ mo

billed monthly

VFX and animation studios with several farms and vendors.

  • —40 TB hot cache
  • —Up to 48 agents
  • —3 regions
  • —SSO (SAML 2.0, OIDC), audit log
  • —Deadline, OpenCue, ShotGrid, ftrack integrations
  • —8×5 engineering support, 4-hour response

Enterprise

from$2,650/ mo

annual contract

Multi-site studios and render service providers.

  • —Cache nodes in your data centre (on-prem)
  • —Unlimited agents
  • —Keys in your own HSM
  • —99.95% SLA, 24/7 on-call engineer
  • —Security audit document pack
Add-ons
In-region trafficfree

Prices exclude VAT. Invoiced under a signed agreement. During the closed beta, plans are available to invited studios only.

04 / FAQ

Common questions

Something else? Write to sales@rendercache.ru.

How is RenderCache different from shared file storage?+

Storage copies whole files. RenderCache understands the structure of VDB, Alembic and UDIM textures, sends only the changed parts and keeps a hot copy next to the farm.

Can RenderCache read our caches?+

No. Data is encrypted on your machines before it is sent, and project keys never leave your perimeter. Cache nodes store ciphertext only.

What happens if the link drops mid-transfer?+

Nothing is lost: upload sessions are kept for 7 days. Once the link is back, the agent continues from the last verified chunk.

Can we run cache nodes on our own hardware?+

Yes. On the Enterprise plan, cache nodes and the control server run in your data centre. You need Linux x86_64, an NVMe array and a 10 Gbit/s+ uplink.

How do we join the closed beta?+

Send a request. We get in touch with your studio, agree on a 21-day pilot and issue accounts to your workspace administrators.

Closed beta

Connect your first farm in a day

21-day pilot, up to 8 agents, with an engineer on call.

Sign in

Request access

Leave your details — an account manager will get back to you within one business day to agree on a pilot and prepare the agreement.