Hypervisor Kernel Scheduler for Low-Latency Telco Workloads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual machine (VM) workloads in telecommunication companies face high latency and jitter due to CPU contention in virtualized systems, as traditional methods of granting exclusive affinity to physical CPUs do not adequately address the needs of network and storage stacks, which are highly input/output (I/O) intensive and require low-latency responses.
Innovation Solution
A method where a hypervisor kernel scheduler identifies system contexts servicing I/O intensive workloads and grants them exclusive affinity to physical CPUs, migrating tasks and directing interrupt processing to other CPUs, while dynamically adjusting CPU reservations to ensure low-latency and fast response times, even under bursty workloads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If virtual CPUs are granted exclusive affinity to physical CPUs, then scheduling delay is minimized and latency is reduced, but CPU contention in the networking and storage stacks still causes jitter and latency
Solution Approach 1:
The patent segments the CPU resources by creating separate run queues for virtual CPU contexts and system service contexts. This segmentation allows independent scheduling of I/O intensive workloads away from general system tasks, eliminating the performance degradation caused by CPU contention in networking and storage stacks while maintaining the low scheduling delay benefits of exclusive affinity.
2Productivity
If system services are implemented outside traditional virtual CPU contexts, then they can service I/O intensive workloads, but they experience CPU contention that adds jitter and latency
Solution Approach 1:
The patent introduces a hypervisor-level intermediary mechanism that intercepts I/O requests from virtual CPU contexts and routes them to dedicated system service run queues. This intermediary layer enables system services to handle I/O intensive workloads while being scheduled on separate physical CPUs, thus maintaining high I/O service capability while eliminating the jitter and latency caused by CPU contention.
3Speed
If physical CPUs are dedicated to virtual CPUs, then latency-sensitive VMs achieve fast response times, but I/O intensive workloads still suffer from shared resource contention
Solution Approach 1:
The patent adds a new dimension to CPU scheduling by implementing a hierarchical run queue structure at the hypervisor level. This dimensional addition allows the system to maintain exclusive CPU affinity for latency-sensitive VMs while simultaneously creating dedicated scheduling paths for I/O intensive workloads through separate system service run queues, thereby achieving both fast response times and reliable packet delivery without drops.
Data Source
AI summary
An approach for providing low-latency and fast response times required by virtual machine (VM) telecommunication (telco) workloads is disclosed. In one embodiment, a kernel scheduler identifies hypervisor system contexts that are outside of traditional virtual central processing unit (CPU) contexts and servicing telco workloads. The scheduler grants each such context exclusive affinity to a respective physical CPU, up to a configurable maximum number of such grants. The scheduler also increases the CPU reservations in a system resource pool for the hypervisor by one CPU, indicating that there is one fewer physical CPU on which VMs may be placed or moved. The exclusive affinities and CPU reservations may be dynamically adjusted as, e.g., new latency-sensitive VMs are powered on with virtual CPUs that need to be granted exclusive affinity to physical CPUs or contexts no longer relevant to telco workloads are demoted to not have exclusive affinity to physical CPUs.