I/O Performance Modelling & Evaluation - giving your company the competitive edge

Abstract

I /O Subsystem performance is one of the key components that impact the business operations of today's organisations. The objective of this paper is to provide a detailed performance evaluation of various storage architectures and strategies that can be used to improve performance. The evaluation is based on a modeled peak on-line database (ADABAS ) workload which provided initial response time estimates. The paper then examines the actual performance improvements that were achieved by moving the workload to devices that support the following architectures: 1. a traditional single stage cache, 2. a managed single stage cache, 3. a two stage cache, 4. and a three stage cache architecture. The effects of multi-stage cache are investigated specifically in relation to cache hit ratios and the performance implications of a second level segmented cache. Cache management techniques are reviewed in relation to the performance improvement that was achieved. Finally the paper reviews the positioning and management of various performance availability, enhanced functions and reliability options based on the business requirements of the organisation. Note: It should be noted that a/though every effort has been made to maintain consistency across the workload comparisons, there are a large number of variables that will effect the overall performance of the database and I/O subsystems. Thus Edgars Stores Ltd. specifically state that the results do not represent any formal type of performance guarantee.