Pairwise Workload Migration for Load Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Standard deviation-based load balancing approaches often fail to identify load balancing opportunities in clusters with outliers, as they may not trigger balancing actions when standard deviation remains below a threshold, leading to prolonged high resource utilization in these hosts.
Innovation Solution
Implementing a pairwise comparison method to determine resource utilization differences between hosts, where workloads from more loaded hosts are migrated to less loaded hosts if the difference exceeds a threshold, ensuring balanced resource utilization without overloading or underloading the target hosts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard deviation-based load balancing is used, then load balancing can be triggered when there is significant variation in resource utilization, but load balancing opportunities for outlier hosts are missed when standard deviation remains below the threshold
Solution Approach 1:
The patent segments the load balancing detection process into multiple comparison levels: first comparing each host against the cluster average (standard deviation approach), then performing pairwise comparisons between hosts with significant differences. This segmentation allows detection of both general imbalances and specific outlier cases that would be missed by a single threshold approach.
Solution Approach 2:
The patent adds a new dimension to load balancing detection by introducing pairwise comparison as a secondary detection mechanism. Instead of relying solely on the single dimension of standard deviation from the mean, the system adds pairwise difference comparison as an additional detection dimension, enabling identification of outliers even when overall cluster variation is low.
2Reliability
If pairwise comparison is implemented to detect all load balancing opportunities, then outlier hosts can be identified, but the complexity of the load balancing algorithm increases
Solution Approach 1:
The patent applies preliminary action by first calculating the standard deviation and identifying hosts that exceed the threshold before performing pairwise comparisons. This preliminary filtering reduces the number of pairwise comparisons needed, as only hosts near the threshold or potential outliers require detailed pairwise analysis, not all hosts in the cluster.
Solution Approach 2:
The patent applies local quality by performing pairwise comparisons selectively only for hosts that are near the threshold or identified as potential outliers, rather than performing exhaustive pairwise comparisons among all hosts. This localized approach reduces computational complexity while maintaining reliability in identifying load balancing opportunities.
Data Source
AI summary
Load balancing across hosts in a computer system is triggered based on pairwise comparisons of resource utilization at different host. A method for load balancing across hosts includes the steps of determining a resource utilization difference between first and second hosts, wherein the first host has a higher resource utilization than the second host, comparing the resource utilization difference against a threshold difference, and upon determining that the resource utilization difference exceeds the threshold difference, selecting a workload executing in the first host for migration to the second host.