Capacity Planning via Workload-Resource Utilization Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing capacity planning techniques for computer systems are inadequate as they fail to accurately predict resource utilization due to non-linear and unpredictable relationships between response time, throughput, and resource utilization, often leading to over-provisioning or under-provisioning, and neglect software resources and the complexity of transactional applications.
Innovation Solution
Capacity planning is performed by determining resource utilization as a function of workload, specifically analyzing different request types and their impact on hardware and software resources, using data analysis and regression to predict resource usage and identify bottlenecks, thereby ensuring appropriate resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing capacity planning techniques use predictive models of response time and throughput, then capacity planning can be performed, but the predictions are inaccurate due to non-linear and unpredictable relationships
Solution Approach 1:
The patent changes the parameters being measured from non-linear metrics (response time, throughput) to linear metrics (resource utilization). By focusing on resource utilization as the key parameter and establishing linear relationships between workload characteristics and utilization, the system achieves more accurate and predictable capacity planning without relying on unpredictable non-linear models.
2Quantity of substance
If capacity planning focuses only on hardware resources, then hardware capacity can be determined, but software resources are neglected leading to incomplete capacity planning
Solution Approach 1:
The patent applies universality by creating a unified capacity planning approach that simultaneously handles both hardware and software resources. The system models resource utilization for multiple resource types (CPU, memory, disk, software licenses, database connections) using a common framework based on workload characteristics, making the capacity planning process versatile across different resource categories.
3Device complexity
If capacity planning does not consider inter-dependencies of composite applications, then individual component capacity can be determined, but the impact of component faults on dependent components is not understood
Solution Approach 1:
The patent segments the composite application into individual components and models the resource utilization of each component separately based on workload characteristics. By breaking down the complex application into manageable segments and analyzing their inter-dependencies, the system can identify how faults in one component affect dependent components, enabling more reliable capacity planning for the entire system.
Data Source
AI summary
Capacity planning based on resource utilization as a function of workload is disclosed. The workload may include different types of requests such as login requests, requests to visit web pages, requests to purchase an item on an online shopping site, etc. In one embodiment, data is determined for each of a plurality of workloads. The data includes characteristics of a workload and resource utilization due at least in part processing that workload. Based on the data, utilization of each of the resources as a function of workload characteristics is estimated. Further, based on the estimated resource utilization, workload characteristics that are expected to cause each respective resource to reach a certain level are predicted. That level could be 100 percent utilization, but could be another level. Capacity planning is performed based on the workload characteristics that are expected to cause each respective resource to reach a certain level.