Transaction Workload Modeling for Web Server Performance Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current web-server log analysis focuses on raw data at the page or request level, leading to inaccurate performance metrics and manual interpretation challenges, failing to effectively compute throughput in terms of transactions per second and responsiveness at the transaction level, resulting in lost critical information and inaccuracies in performance test strategies.
Innovation Solution
A method and system that capture performance metrics from production web server logs at the transaction level by creating a transaction log from activity logs with date, logged time, session identification number, and destination address, determining transaction metrics like average response time and throughput, and using these to generate a workload model for testing critical application performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If web-server log analysis focuses on raw data at page or request level, then data extraction is straightforward, but measurement precision of performance metrics deteriorates
Solution Approach 1:
The patent segments the analysis process into two distinct levels: raw data extraction at page/request level, and processed analysis at transaction level. By dividing the log data into individual requests and then grouping them into logical transactions, the system maintains ease of data extraction while achieving precise transaction-level performance metrics through structured aggregation and correlation of raw data elements.
2Device complexity
If log analysis views each entry as isolated, then analysis process is simple, but loss of information increases
Solution Approach 1:
The patent merges multiple isolated log entries into cohesive transaction structures by correlating requests through session identification and logical transaction boundaries. Individual page requests, which appear as separate isolated entries, are combined and grouped into meaningful business transactions, preserving critical information about user workflows and application performance while maintaining analytical simplicity through automated correlation rules.
3Measurement precision
If throughput is measured in bytes per second at page level, then measurement is straightforward, but productivity of performance testing deteriorates
Solution Approach 1:
The patent transforms the throughput measurement parameter from low-level technical metrics (bytes per second at page level) to high-level business metrics (transactions per second). By changing the unit and context of measurement to align with business transaction definitions, the system maintains straightforward measurement capabilities while dramatically improving performance testing productivity through metrics that directly reflect application performance from a user and business perspective.
Data Source
AI summary
A method for creating workload model to test performance of a critical application in a data processing network (112) is disclosed. The method includes receiving (202), at one of the plurality of data processing units (114a), an activity log with a plurality of entities from the data processing network (112). The method further includes creating (204) a transaction log of all transactions conducted using the activity log. Further the method determines (206) a plurality of transaction metrics using the transaction log, wherein the plurality of transaction metrics comprises an average transaction response time, a transaction throughput etc. The method further creates a workload model to test the critical performance of an application in the data processing network (112) based on the plurality of transaction metrics and provide effective performance test strategies.