NODE: EDGE-01 [PROFILE] INGRESS: QUIC / TLS 1.3RTT: 1.4 ms UPTIME: 99.994% CLUSTER PROFILE: GRID-01

KVALV GRID OS v4.8 GLOBAL MEDIA GRID

8K Post-Production Over Global Optical Fabric

Block-streamed media. Cloud post-production. No waiting for the entire file.

Bring 8K originals straight to the timeline. Move live signals across borders. Connect editors, colorists and render nodes worldwide through one modular media grid.

Open WorkspaceConfigure infrastructure
BLOCK READ / TARGET
40 Gbps
CHUNK SIZE
4 MiB
REVIEW / TARGET
<100 ms

Performance depends on the allocated port, source codec and end-to-end network path.

GRID-01 / GRID FABRICFABRIC READY
KVALV blade rack and optical fabric topology Four blade modules connect to an optical backplane: V-01 Vault with NVMe cassettes, R-02 Relay, C-03 Compute and S-04 Shield. Telemetry indicators and throughput traces show each circuit. Select a mode below to highlight its module and data path. X / 000GRID-01X / 640GRID / v4.8REV 4.8 / 04 UNITS BLADE RACK / OPTICAL FABRICTHROUGHPUT TRACES / 60 s V-01 VAULTNVMe / PCIe 4.0V-01 / GRID-01 N0N1N2N3N4N5 R-02 RELAYSRT / QUICR-02 / GRID-01 TX0TX1TX2RX0RX1RX2 C-03 COMPUTECPU / GPUC-03 / GRID-01 C0C1C2G0G1G2 S-04 SHIELDHSM / WORMS-04 / GRID-01 K0K1K2K3K4K5 FULL-DUPLEX OPTICAL BACKPLANE
BITRATE
40 Gbps
IOPS
1,200,000
CHUNK
4 MiB

Built for the formats you actually work with.

FROM CAMERA ORIGINAL TO FINAL DELIVERY
TRANSPORT
  • HTTP/3 QUIC
  • SRT over TLS / gateway
  • WebRTC WHIP/WHEP
  • S3 Multi-Part Parallel
  • FUSE / SMB3 Block Stream
FORMATS
  • ARRIRAW
  • RED R3D
  • Apple ProRes 4444 XQ
  • OpenEXR 32-bit
  • Dolby Atmos Master
  • SMPTE DCP

Protocol names describe integration paths. Native SRT uses its own transport security; TLS terminates at the gateway where configured.

01 / THE MODULAR ADVANTAGE

Modular Infrastructure Built for High-Bitrate Pipelines

Allocate storage, transport, compute and protection independently. Each module handles a defined part of the production pipeline.

V-01VAULT MODULE

Block-streamed access to original media

An NVMe-backed media pool presents remote originals to your NLE. Only the blocks needed for the current playhead are streamed and cached by the workstation client.

  • Up to 40 Gbps per allocated workstation port
  • 4 MiB chunks with sequential read-ahead
  • 8K RAW timelines over FUSE / SMB3 integration
Explore the media pool
R-02RELAY MODULE

Low-latency ingest and remote review

Route live ingest and review feeds through a dedicated media edge. SRT, QUIC and WebTransport paths serve broadcast contribution and responsive client sessions.

  • Sub-100 ms remote grading target for DaVinci Resolve
  • WHIP ingest and WHEP review integration
  • Per-session bandwidth allocation and jitter budget
Size your live pipeline
C-03COMPUTE MODULE

Shared-source proxy and render workers

Fan out incoming media to proxy and transcode workers. Keep render nodes synchronized to the same source revisions, without copying the entire production to each machine.

  • On-ingest proxy generation and codec presets
  • Shared source blocks across render workers
  • Queue-aware allocation for burst workloads
Connect your pipeline
S-04SHIELD MODULE

Client-key encryption with retention locks

Client-side chunk encryption keeps plaintext outside the storage perimeter. Immutable WORM snapshots preserve selected revisions for an agreed retention window.

  • Cryptographic verification of every 4 MiB chunk
  • Client-controlled keys and recovery responsibility
  • Retention locks defined at allocation time

02 / LIVE STUDIO WORKSPACE

A Shared Workspace for Every Production Asset

Inspect camera originals, review feeds, image sequences and spatial masters from one media pool.

KVALV / GRID-01 / MEDIA POOLSNAPSHOT: GRID-01 MEDIA POOL

8 objects 6.00 TB

Metadata catalog for the GRID-01 media pool

Live inputs show 0 GB until recorded.Capacity units: 1 TB = 1,000 GB.

Infrastructure configurator

Size your NVMe pool, uplink and cache nodes against a single monthly budget.

Quick tier presets
TB
Contract term

8 hours cover one shift. The 16-hour dual shift includes priority traffic. Continuous broadcast includes a guaranteed unlimited buffer.

4

Up to 4 included. Each additional node costs €45/month.

Protection & auto-backup

Monthly allocation

€900
Effective rate per TB/month
€18.00
Annual contract savings
€0

Monthly costs before discount

Storage pool
€900
Auto-backup & replica
€0
Network & ingest
€0
Compute cache nodes
€0

Engineering estimates

Raw provisioned capacity
62.5 TB
Peak pool read speed
4 GB/s
Concurrent 8K RED R3D streams
5
Concurrent 8K ProRes 4444 streams
2
Available uplink headroom
20%

Sizing model: 1 GB/s of pool reads per node. Stream counts assume either 2 Gbps per R3D stream or 4 Gbps per ProRes stream, limited by pool and uplink. Headroom assumes one R3D stream per node.

Discuss this allocation

Engineering estimates, not a binding quote or a capacity reservation. Final resource availability and pricing require a confirmed allocation.

04 / FROM RACK TO TIMELINE

Connect Your Workstation to the Media Pool

  1. Authenticate at the edge.

    Provision a client certificate and hardware-backed identity. Choose an mTLS or WireGuard access path; media transport uses the allocated QUIC endpoint.

  2. Mount the cloud pool.

    Attach the allocated pool to your workstation through the supplied client. Define a local cache budget and assign project-scoped access.

  3. Stream source blocks into your timeline

    Open the source in your NLE or render pipeline. Read-ahead brings upcoming blocks into the client cache while proxy workers prepare lighter review streams.

KVALV / CONNECTION PROFILE

Illustrative mount command for a provisioned KVALV client

KVALV_REGION='ASSIGNED_REGION'
KVALV_POOL='Feature_Film_8K_Master'
KVALV_MOUNT_POINT='/Volumes/KVALV'
KVALV_CACHE_GIB='128'
KVALV_IDENTITY='hardware-backed'
KVALV_ACCESS='read-only'

kvalv-mount --profile /etc/kvalv/grid01.conf --verify-mtls

Example result after a successful mount

[OK] mTLS verified [OK] GRID-01 / Feature_Film_8K_Master [OK] /Volumes/KVALV · read-only · cache 128 GiB
CONNECTION PROFILE

Examples require the client, certificate and connection profile supplied with your allocation. Substitute assigned parameters before use. No connection is initiated by this page.

05 / ALLOCATION TIERS

Capacity That Scales with Your Studio

A focused edit team, a distributed production or a dedicated facility. Reserve the resource envelope that fits the work.

01 / PODSMALL TEAMS

Studio Pod

A compact shared media pool for editors and colorists working on one production at a time.

From €900/ month

  • 50 TB usable capacity · €18/TB/month
  • 10 Gbps dedicated uplink
  • 4 active cache nodes included
  • Standard dual-parity protection
  • Scheduled proxy processing window
Request Studio Pod
02 / GRIDMULTI-TEAM PRODUCTION

Production Grid

A coordinated storage, compute and relay allocation for parallel post-production and live contribution.

From €3,200/ month

  • 200 TB usable capacity · €16/TB/month
  • 10 Gbps dedicated uplink
  • 4 active cache nodes included
  • Vault, Relay and Compute resource pools
  • Optional Shield retention policy
Request Production Grid
03 / SOVEREIGNDEDICATED INFRASTRUCTURE

Dedicated Sovereign Cluster

Dedicated infrastructure for distributed teams that need defined tenancy, custom network paths and their own retention policy.

Custom allocation

  • Dedicated nodes and capacity planning
  • Unmetered 40 / 100 Gbps port options
  • BGP / private peering by agreement
  • Hardware identity and WORM policy design
  • Contract-defined support and availability
Plan a dedicated cluster

Studio Pod and Production Grid match the configurator's base presets. Contract term, protection, uplink, ingest window and extra nodes change the estimate. VAT and availability are confirmed in the quote.

06 / BEFORE YOU CONNECT

The engineering details matter.

Understand the connection model, the data boundary and the responsibilities on each side of the grid.

Why maintain persistent QUIC / TLS connections?

A media client can reuse a secure session across many block requests, avoiding a new connection handshake for every chunk. QUIC supports independent streams and connection migration where configured. This page opens no media connection and does not measure your network.

How can I edit without downloading the entire file?

The mounted client translates file reads into block requests, caches the returned chunks and prefetches nearby data. Your NLE sees a file path while the client retrieves the ranges it needs. Random seeks, codec structure, cache size and link capacity all affect playback; an initial buffer may still be required.

How is data residency defined across regions?

The service agreement defines the regions used for primary media, replicas and backups. It also specifies support access, subprocessors and cross-border transfers. Choose an allocation that matches your production workflow and data-residency requirements.

Why is ordinary password sign-in unavailable?

The intended access model uses provisioned client identities rather than a public password endpoint. mTLS certificates and FIDO2 authenticators serve different roles and require their respective client integrations. The browser console on this page performs only a local interaction sequence; it cannot authenticate to a node.

How do Zero-Knowledge storage and WORM work together?

The client encrypts media before upload; a retention policy can then lock the stored ciphertext and selected snapshots against modification. Storage operators should not possess the decryption keys. Transcoding plaintext requires a separately authorized trusted processing path. WORM cannot recover lost client keys, and its lock period must be chosen carefully.

Provisioning & Network Inquiry

GRID-01

Generate a pre-validated resource specification for your studio’s procurement and network engineering team.

Include capacity, codecs, concurrency and delivery requirements. The specification is generated locally for procurement review; capacity is reserved after commercial confirmation.

ASSET INSPECTOR / BLOCK MONITOR

FEATURE_A001_C014.arriraw

Local frame and stream-profile preview. No source media is downloaded or played.

Signal monitor and vectorscopeA locally animated signal trace with frame guides, a luminance waveform and a circular color vectorscope. Y / RGB PARADEVECTORSCOPENVME-01 / FRAME MONITOR
RAW / 480 MbpsPAUSED
01:14:08:2224 fps

KVALV / GRID-01

Node Console

Inspect the hardware-identity access flow. This local console does not contact a node, read certificates or request access to an authenticator.

Use an invented value only. Never enter a real access token or password.

SECURITY / PRIVACY / NETWORK

Security, Privacy & Network Policy

This notice describes the public interface and the intended allocation model. Binding service terms, data-processing responsibilities and technical controls must be confirmed in the signed service agreement.

01 / Public interface and local state

The interface uses no cookies, analytics, tracking pixels or browser storage. Search, filters, calculators and inquiry packages are held in page memory. It makes no background API, media or authentication requests. Loading the page and its same-origin assets still involves ordinary HTTP requests; hosting access logs are governed separately.

02 / Client encryption and key ownership

In the intended Zero-Knowledge storage model, a provisioned client encrypts media chunks before transmission. The customer controls the decryption keys and their recovery process. This page does not implement media encryption, hold key material or validate a production deployment. Plaintext processing requires an explicitly trusted compute environment and cannot be assumed to preserve a storage-only Zero-Knowledge boundary.

03 / Integrity, retention and WORM

Checksums can identify a change in the covered bytes; they do not prove identity or provide a digital signature. A production retention lock must define its scope, retention period, authorized roles and deletion behavior. Immutable ciphertext does not replace key backup or disaster recovery. Source media integrity is not checked by this website.

04 / Data residency and processing terms

For each allocation, the service agreement specifies the regions used for primary media, replicas and backups. It also identifies the operator, controller and processor roles, subprocessors, retention, support access and international transfers. Applicable requirements depend on the agreed locations and processing activities. No certification or regulatory-compliance claim is made by this page.

05 / Acceptable use of network capacity

Allocated bandwidth is intended for authorized production, delivery and review workloads. Do not use it for attacks, unsolicited bulk traffic, unauthorized scanning, unlawful distribution or interference with other tenants. An unmetered allocation remains bounded by the agreed physical port rate, routing policy and fair-use terms. Incident measures and restoration procedures belong in the service agreement.

06 / Identity and operational access

Use individually provisioned identities and the minimum required project permissions. mTLS authenticates a TLS client using a certificate; FIDO2 authenticates through a separate relying-party integration. Do not paste credentials into this public page. The local console does not verify tokens, connect to a client daemon or test a hardware authenticator.

07 / Requests and clipboard handling

An inquiry package remains local until you choose to copy and transfer it. Its SHA-256 checksum covers the serialized payload only; the package is unsigned and not registered with the service. Clipboard actions require your interaction and may be restricted by the browser. Copied content may remain in the operating system clipboard after the page closes. Review the package before handing it to an authorized recipient.

08 / Security and abuse reports

Use the Network security / abuse inquiry category to prepare a report with timestamps, affected public endpoints and a concise description. Exclude credentials, private media and unnecessary personal information. Transfer it through the operator contact channel specified in your contract; preparing a local package does not notify an operator.