Get the newsletter
IBM · Memory & storage · Filed Nov 8, 2024 · USPTO published May 14, 2026 · May 14 release

IBM Patents a Resilient Network Cache System for Virtual Machine Environments

When a virtual I/O server crashes in a busy data center, every virtual machine depending on it can grind to a halt. IBM's new patent describes a system that keeps the cache humming even when parts of the infrastructure fail.

IBM Patent: Network Cache Resiliency for Virtual Machines, figure from US 2026/0133910 A1
Figure from the official USPTO publication.
Publication number
US 2026/0133910 A1(US20260133910A1)
Applicant
INTERNATIONAL BUSINESS MACHINES CORPORATION
Filed · Published
Nov 8, 2024 · May 14, 2026
Inventors
Veeresh Jumanal, Rizwan Sheikh Abdulla, Shivarudrappa Satyanaik, Ravikishore Krishnamurthy
Status when we published
Company responded; back with the examiner
Document
20 claims
US classification
711/145
Examiner
FAAL, BABOUCARR (Art Unit 2138)

What IBM's virtual cache pooling system actually does

Imagine a busy office building where every floor shares a single copy machine down the hall. If the one courier who normally runs copies for your floor gets sick, nothing moves — even though the copier itself is fine. That's roughly the problem IBM is solving here, but for data centers.

In this patent, IBM describes a way to take a network-attached cache device — essentially fast temporary storage sitting on a remote server — and share it intelligently across many virtual machines (VMs). Each VM gets its own slice of that cache, and multiple virtual I/O servers (the couriers in our analogy) are all wired up to handle requests. If one server fails, the others automatically pick up the slack.

The practical result: your workloads keep running and your cache stays accessible even during partial infrastructure failures. It's the kind of unglamorous reliability work that enterprise IT teams quietly depend on every day.

How IBM slices and load-balances a remote cache pool

The patent describes a three-part architecture running on a host system that manages multiple virtual machines (VMs — software-based computers running on shared hardware).

First, the host connects to a network-accessible cache device hosted on a remote system. Think of this as a high-speed scratch pad that lives across the network rather than directly on the server. The host then exposes that remote cache to a set of virtual I/O servers (vIOS) — lightweight software intermediaries that handle storage traffic on behalf of VMs.

Second, the system virtualizes the remote cache into discrete cache partitions within a logical cache pool. Each partition is assigned to exactly one VM, so VM-A's cached data is isolated from VM-B's — preventing cross-contamination while still sharing the underlying hardware.

Third — and this is the core claim — a cache access management layer handles every incoming cache request by:

  • Load balancing requests across multiple vIOS instances so no single server becomes a bottleneck
  • Failover recovery: if one vIOS fails, the management layer reroutes traffic to surviving vIOS instances transparently
  • Maintaining per-VM partition assignments throughout both normal operation and failure events

The result is a cache infrastructure that is both horizontally scalable and fault-tolerant without requiring VMs to be aware of the underlying complexity.

What this means for enterprise cloud reliability

Enterprise data centers run enormous numbers of VMs, and cache performance is often the difference between a snappy application and a sluggish one. Today, many cache architectures are tightly coupled to individual servers, meaning a single hardware or software failure can cascade into widespread slowdowns. IBM's approach decouples the cache resource from any single access path, making it far harder for a localized failure to ripple outward.

For IBM's Power Systems and cloud infrastructure customers — think large banks, telcos, and government agencies — this kind of resiliency is table stakes. The patent fits naturally into IBM's existing virtualization stack and could tighten integration with offerings like PowerVM or IBM Cloud. You probably won't notice it directly, but your applications running on IBM infrastructure would be more likely to stay fast when something underneath them breaks.

Source: US 2026/0133910 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time IBM patent news breaks.

There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.