Correlated Workload Clustering to Reduce Stranded Server Resources
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The bin packing problem in data centers, where workloads are inefficiently allocated due to computational complexity and resource strandedness, particularly in on-demand code execution systems with varying resource utilization.
Innovation Solution
A load balancer utilizes clustering algorithms to characterize expected resource usage of workloads and servers, reducing computational complexity by maintaining information about workload clusters rather than individual loads, and routes requests based on correlation or anti-correlation of resource usage shapes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional bin packing algorithms are used to allocate workloads, then workload allocation can be achieved, but computational complexity becomes too high and resource utilization is inefficient
Solution Approach 1:
The patent segments workloads into clusters based on their resource usage characteristics. Instead of treating each workload individually, similar workloads are grouped together, reducing the complexity of allocation decisions while maintaining efficiency. This segmentation allows the system to make faster allocation decisions without sacrificing optimization quality.
Solution Approach 2:
The patent changes the parameters used for workload allocation from detailed individual workload characteristics to cluster-level aggregated parameters. By transforming individual workload attributes into cluster representations, the system reduces computational complexity while preserving the essential resource usage patterns needed for efficient allocation.
2Productivity
If workloads are allocated to minimize the number of devices used, then resource efficiency improves, but stranded resources occur when workload shapes do not match server resource shapes
Solution Approach 1:
The patent applies local quality by matching the resource usage shape of specific workload clusters with the corresponding resource shape of suitable servers. Instead of uniform allocation, the system tailors workload placement to the local characteristics of both workloads and servers, ensuring that workloads with similar resource patterns are placed on servers with matching resource availability, thereby minimizing stranded resources.
Solution Approach 2:
The patent creates a simplified representation or 'copy' of the resource usage shape for each workload cluster. This copied shape information is then used to quickly match workloads with suitable servers without requiring detailed analysis of individual workload requirements, enabling efficient allocation while avoiding stranded resources through shape-based compatibility matching.
3Measurement precision
If detailed information about individual workloads is maintained for precise allocation, then allocation accuracy improves, but system complexity and memory requirements increase
Solution Approach 1:
The patent merges information about individual workloads by grouping them into clusters with similar resource usage characteristics. Instead of maintaining separate detailed information for each workload, the system consolidates their attributes into cluster-level representations, reducing information management complexity while preserving sufficient accuracy for effective allocation decisions.
Data Source
AI summary
Systems and methods are described for allocating requests to implement new workloads within a set of servers. Each server can have a given proportion of various resources, based on a hardware configuration of the server. Placing multiple workloads that do not match that proportion can result in stranded resources, which are unused but cannot be used due to a lack of other required resources. Embodiments of the present disclosure include load balancer that routes workload requests based on the proportion of resources expected to be used by the workload, by selecting a target server based on whether the target server hosts other workloads that are correlated or anti-correlated with the requested workload. To reduce maintained state information, the load balancer can characterize workloads in terms of clusters, rather than maintaining information as to individual workloads.