Local Performance Index for Multi-Tier Server Workload Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing workload management approaches, particularly goal-oriented methods like IBM Workload Manager (WLM) for z/OS, are limited in managing multi-tiered transaction-based workloads as they only consider single-hop transactions, failing to effectively estimate the impact of resource changes on overall end-to-end performance in multi-hop environments and requiring large data transfers for monitoring.
Innovation Solution
A method and apparatus for calculating a local performance index for a particular server in a multi-tiered environment by scaling the contribution of that server's response time within its tier, allowing for efficient monitoring of all transactions without extensive data transfer, and projecting changes in performance due to resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If single-hop transaction management is used (IBM WLM for z/OS), then transaction response time can be measured between two points within the same process, but it cannot effectively manage multi-tiered transaction-based workloads or estimate the impact of resource changes on overall end-to-end performance
Solution Approach 1:
The patent segments the end-to-end transaction performance into individual server contributions by introducing the local performance index metric. Each server's performance is measured and weighted independently based on its contribution to the overall transaction, allowing multi-tiered workloads to be managed through composition of individual server metrics rather than treating the entire transaction as a single unit.
2Measurement precision
If complete transaction monitoring is implemented across all servers in a multi-tiered environment, then accurate end-to-end performance measurement is achieved, but large amounts of data transfer are required for monitoring
Solution Approach 1:
The patent extracts only the necessary performance data from each server (local performance index and contribution weight) rather than transferring complete transaction logs. This extraction approach maintains measurement precision by capturing the essential performance metrics while minimizing data transfer overhead by excluding redundant information.
3Measurement precision
If average response time is used as the performance metric, then overall transaction performance can be measured, but the contribution of individual servers cannot be determined
Solution Approach 1:
The patent applies local quality by defining the local performance index as a server-specific metric that captures each server's individual contribution to transaction performance. Each server is evaluated with its own performance characteristics and weighting factor, allowing precise attribution of performance to individual servers while maintaining relatively simple calculation procedures.
Data Source
AI summary
A method and apparatus for obtaining a local performance measure for a particular server in a particular tier in a transaction environment in which transactions pass through multiple tiers with multiple servers at each tier. The contribution from the particular server to the total end-to-end response time for a set of transactions is scaled by the ratio of transactions passing through the particular tier to transactions passing through the particular server to obtain a scaled contribution from the particular tier. This is added to the contribution from outside the particular tier to obtain a modified total end-to-end response time from the perspective of the particular server. The modified total end-to-end response time is divided by the number of transactions in the set to obtain a modified average end-to-end response time from the perspective of the particular server, which is used to control allocation of resources to the server.