Home Multikernel Private Cloud Multikernel Sandbox Multikernel LiveUpdate ARM Platform RISC-V Platform OEM & Embedded SaaS & Database Clouds Technology FAQ Getting Started Blog About 中文 GitHub Schedule a Demo
OEM & Embedded

Ship Multikernel Inside Your Appliance

Built for storage and appliance vendors who embed an operating system and ship it to their customers. Multikernel splits one controller into independent kernel fault domains: a driver fault becomes a restartable component instead of a controller-wide panic, and a kernel patch no longer triggers a failover event.

Sub-Node Fault Domains

In a monolithic controller OS, a single driver fault panics the entire node, triggering controller failover, peer degradation, and a DRAM cache re-mirror. The split-kernel architecture reduces the fault domain to one layer of the I/O stack: kernel boundaries follow the data path, and each layer fails and recovers independently.

A Storage Controller · Example Kernel Fault-Domain Split
Front-end kernel
FC / Ethernet · protocol termination
Data kernel
cache · data protection · data services
Back-end kernel
SAS / NVMe drivers

A driver fault restarts one kernel. No controller failover, no cache re-mirror event.

Component-Level Fault Recovery

A failed driver kernel restarts in place while the data kernel continues serving from cache. The impact is confined to one layer of the I/O stack rather than the node covered by your availability SLA.

Structural Tail-Latency Isolation

Rebuild, resync, and background scrubbing run in their own kernel on their own cores. p99.9 latency is protected by architectural isolation rather than by tuning that degrades under load.

Per-Layer Kernel Specialization

An interrupt-free data kernel, a polling device kernel, and smaller synchronization domains on many-core controllers. Each kernel is configured for a single role.

LiveUpdate as the Entry Point

LiveUpdate integrates into an existing storage OS without architectural changes. It deploys standalone, per node, inside the operating system you already ship. Adoption is sequenced to limit risk: LiveUpdate proves itself inside your current OS first, and the fault-domain split follows only when evaluation results support it.

Kernel Patches Without Failover

The patched kernel boots alongside the running one; the data path migrates incrementally, with instant rollback available throughout. No peer degradation, no cache re-mirror, and no maintenance window on single-controller systems.

CVE Management for Appliances

Security backports delivered under SLA, an errata feed that integrates with your release process, and a ready-to-apply LiveUpdate image for every fix. Field systems stay patched without reboot-driven support escalations.

Sub-Second Crash Failover

A standby kernel monitors the primary and takes over its devices in under a second, removing kernel panics as a driving factor in your availability model.

Licensing and Redistribution

Multikernel is licensed under GPL-2.0 and remains fully open source. The OEM subscription covers what commercial redistribution requires: the rights and build infrastructure to ship signed Multikernel builds under your brand, backed by SLA-based engineering support behind your support organization.

Open Source Base

Built on upstream Linux under GPL-2.0. Evaluate, prototype, and embed without license negotiation. Joint engineering work is contributed upstream.

OEM Subscription

Priced per shipped controller. One validated kernel stream to track, instead of engineering effort spent chasing upstream kernel churn across your product lines.

Redistribute signed, validated Multikernel builds inside your appliance OS
Your hardware in our validation matrix, every release tested on your controllers
CVE backports under SLA with errata feed and matching LiveUpdate images
Second-line engineering support behind your support organization
Fault-domain design review for your data path

Performance Validated in Evaluation

Performance characteristics are measured during evaluation on your hardware and workloads, not projected from synthetic benchmarks. The architectural benefits are structural: fault containment, tail-latency isolation, and reboot-free maintenance.

Engagement

From Evaluation to Shipping

A three-step engagement model, from proof of concept to a shipping product line.

01

Evaluation

Proof of concept on your controller hardware and I/O stack: LiveUpdate integration first, followed by a fault-domain split, measured on your workloads.

02

Integration

Embed Multikernel into your appliance OS with our engineers working alongside your team: fault-domain boundaries for your data path, device kernel enablement for your HBAs and NICs.

03

Ship and Maintain

Signed build stream, CVE SLA, and second-line support for the life of the product line. Enablement work is contributed upstream with your platform as the reference implementation.