Shared Receive Queue Polling with Dynamic CPU Core Groups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage node architectures face issues with asymmetric core processing loads leading to increased latency and limited scalability in RPC messaging due to core contention and resource sharing, especially in multi-core CPU environments.
Innovation Solution
Dynamically changing groups of CPU cores for polling shared receive queues (SRQs) based on system load thresholds, adjusting the number of core groups and cores per group to optimize latency and throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If each RPC request/reply is associated with specific corresponding cores, then core processing specialization is improved, but latency increases when cores have asymmetric processing loads
Solution Approach 1:
The patent implements dynamic core group configuration where the number of cores assigned to poll SRQs changes based on system load conditions. During low load, fewer cores poll SRQs to reduce contention; during high load, more cores are allocated to handle increased RPC traffic, optimizing both latency and throughput adaptively
2Productivity
If a shared queue is used to receive RPC requests/replies, then core utilization is improved, but resource contention among cores increases
Solution Approach 1:
The system dynamically adjusts the number of polling cores based on system load thresholds. When load is low, fewer cores poll the shared queue to minimize contention; when load increases, more cores are allocated to polling, balancing resource utilization against contention overhead
Solution Approach 2:
The patent changes the parameter of polling core count based on system load conditions. By monitoring load metrics and adjusting the number of active poller cores, the system optimizes the balance between queue drainage efficiency and resource contention
3Productivity
If the number of cores available for RPC messaging increases, then throughput capacity is improved, but scalability is limited by heterogeneous core configurations
Solution Approach 1:
The patent creates a universal core group abstraction that works across heterogeneous multi-core systems. By grouping cores and managing them as a unified resource pool rather than individual units, the solution achieves consistent behavior across different CPU configurations and enables scalable RPC messaging
4Loss of time
If poller cores are highly available for polling SRQs, then RPC processing latency is reduced, but system load handling capacity is limited
Solution Approach 1:
The system implements dynamic scaling of poller core availability based on system load. During low-load periods, a small number of dedicated poller cores maintain low latency for RPC processing. When system load increases, additional cores are allocated to polling duties, increasing throughput capacity while maintaining acceptable latency performance
Data Source
AI summary
Techniques for dynamically changing groups of CPU cores for polling shared receive queues, providing a tradeoff between decreasing latency and increasing throughput of storage nodes using RPC messaging. The techniques include, upon initialization of a storage node, assigning all its CPU cores to the same core group. The techniques include, in response to detecting that its system load has been maintained above a threshold value for a specified time interval, increasing the number of core groups by a predetermined factor, and decreasing the number of cores assigned to each core group by the predetermined factor. The techniques include, in response to detecting that the system load has been maintained at a level less than the threshold value for the specified time interval, decreasing the total number of core groups by the predetermined factor, and increasing the total number of cores assigned to each core group by the predetermined factor.