Home > Technical Papers/Books >

  Predicting The Performance Of 3 Tier Systems. (11-Mar-2002)
 
 
Our Price: Free


Availability: Available for download now
Product Code: 3TIER

Description
 
This white paper describes an approach the Transaction Processing Service Line within Oracle Consulting has developed for estimating the resource utilisation of a 3 tier system on the database server. The first section describes some basic concepts that we used in developing this approach, followed by a discussion of the risks and assumptions used. The second section describes the experimental techniques used to produce performance estimates. The last section describes the tools we used to carry out the investigation The method arose out of a need to answer to basic questions for a customer who was developing a 3 tier, Oracle*Forms / Tuxedo/ Oracle database based billing system. The customer wanted to know:

Will the application architecture scale to meet the customer's throughput requirements within the hardware and software constraints?

Will it scale to meet future, and indeterminate, requirements.

In order to answer these questions we determined the cost of running each Tuxedo service, and then modelled the production workload. We then presented an overall risk assessment, showing the risk that the target server could or could not support the expected workload.

Our approach is a reductionist approach, in which we break down the system into small pieces, and evaluate the behaviour of the pieces in isolation. We then use a model to predict the behaviour of the whole system. Our method does not include the workload generated by applications outside of our test scenarios (batch work for example). This load would need to be factored in if it runs at the same time as the measured workload.

Our approach means that we can model hypothetical workload profiles. This was important for the customer, as the nature of their business is changing due to legislative and market changes. The planning and design of the tests form a large part of this whitepaper. Time spent defining test data and transactions in advance will mean that the actual process of testing runs more smoothly.

The examples used in the paper illustrate points, and are not derived from actual testing.
 


Browse for more products in the same category as this item:

Technical Papers/Books


Essential Performance Forecasting - Presentation RMOUG/IOUG/Collab'06/OOW'06 (October 2006) Free
The Ratio Modeling Technique - Quickly Performing Low Confidence Capacity Predictions (July 2003) Free
Stock Trader Load Generator (2h 19-April-05) Free
Grid Computing—The Next Big Thing (August-2003) Free
OraPub System Monitor (OSM) Toolkit (v12y for Ora 7-11g, 08-Feb-2010) Free
Response Time Graph Template (CPU(1,2) & IO) (1b, 8-Jun-2010) Free
The Effectiveness of Global Temporary Tables (jan-01) Free
Simple Linear Regression Analysis Template (2n 09-July-07) Free