Virtualized cloud and data center environments struggle to apply hypervisor and OS updates without service interruption. Traditional approaches rely on rebooting, live migration across hosts, or iterative copying of memory, all of which introduce downtime, generate significant network traffic, and require spare capacity. As memory sizes and dirtying rates grow, copy-based migration becomes slow and unreliable, while live patching remains limited and depends on the stability of potentially faulty systems. These constraints prevent operators from performing timely security upgrades and maintenance without disrupting running applications.
The invention enables live replacement of hypervisors, operating systems, and intra-host container migration by remapping memory ownership instead of copying memory pages. A hypervisor-resident mWarp table records GVA-to-HPA mappings, allowing a destination VM to access existing memory directly. A thin hyperplexor layer beneath the hypervisor manages multi-level address translations using shadow EPT and HyperFresh tables, enabling seamless switching to a new hypervisor instance after transferring only VCPU and I/O state. For live OS replacement, containers are relocated between VMs on the same host using the same memory-remapping technique. The system integrates with CRIU and adds new syscalls, hypercalls, and virtualization optimizations to support efficient transitions.
• Enables sub second transitions by remapping memory instead of copying it
• Eliminates dependency on network bandwidth during migration
• Reduces downtime for hypervisor and OS updates to milliseconds
• Supports proactive rejuvenation of hypervisors and operating systems
• Avoids reboots and inter host migrations for routine maintenance
• Improves reliability by removing dependence on buggy running systems
• Provides efficient intra host container migration
• Reduces operational complexity in large virtualized environments
• United States 12,093,713 Utility Issued 09/17/2024
• United States US20240403106A1 - Pending
Prototype tested on small-scale virtual machine environments
This technology is available for licensing.
Strong potential for adoption by cloud service providers, data center operators, virtualization platform vendors, and edge computing providers seeking zero-downtime update capabilities, improved system reliability, and reduced operational overhead in large-scale virtualized environments.
Prototype validation results and implementation details available upon request.