As I/O subsystems and storage devices grow in complexity, simple models, while not always accurate enough to calibrate to measurements, are of great value to enhance our understanding of important performance factors of these systems. This session considers simple models of cached I/O, Parallel Access Volumes and hypervolumes, as well as the communication fabric of a Storage Area Network to illustrate this point. An example shows also how simple models can guide our understanding of actual measurement data and help in the subsequent model calibration.

This tutorial was designed to introduce the manager/analyst new to the field to a variety of techniques and methodologies for capacity planning and performance evaluation. In particular: the system life cycle, benchmarking, modeling techniques, workload characterization, and performance management. The primary purpose of this tutorial is to provide an introduction for the novice and an overview for the manager.

Performance management controls of CICS for MVS/ESA and CICS Transaction Server (TS) for OS/390 - z/OS regions can greatly affect the effective use of computer resources, and, as a result, the effective capacity of a CICS complex is impacted. This presentation will focus on the parameters and OS/390 - z/OS environmental factors which affect a CICS region's total demand for processor, real storage, virtual storage, and disk I/O service. The potential risks and benefits associated with the selection of actual and default values will be identified.

Performance and Capacity problems do not occur when a system is conceived or even when the go-live switch is thrown, it is only when users start actually using the system! This session demonstrates the risks of incorrect workload characterisation and presents a number of techniques for soliciting forecasts from business users that can used to derive quality workload characterisations for performance and capacity planning. Also, an example of user workload information hidden within MIS data is shown by a study of weblogs from an Internet Bank.

E-business is maturing; critical e-business systems must be fast, reliable, and effective. The performance of Web-based applications cannot be guaranteed purely through reactive measurement and tuning approaches, it must be managed proactively to meet service levels objectives, using an approach that factors in performance goals throughout all phases of the application life cycle. Based on experiences helping companies meet their e-business performance goals, this tutorial will present a comprehensive framework encompassing the design, development, deployment, and ongoing performance management of e-business applications. The framework addresses site and application design, performance testing, performance monitoring, service level management, alerting, root cause analysis, problem investigation, and problem diagnosis. The approaches suggested are independent of implementation platforms and performance tools, however, some examples of particular technologies and tools will be included."

This tutorial will discuss the basics for UNIX performance tuning using Linux commands for tuning for CPU, memory, Disk and network. Examples will be primarily from SUSE Linux.

Web-based applications are becoming mission-critical for many organizations and their performance has to be closely watched. This paper discusses three important activities in this context: load testing, benchmarking, and application performance management. Best practices for each of these activities are identified. The paper also explains how basic performance results can be used to increase the efficiency of load testing procedures.

This session provides an overview of the InfiniBand Architecture, which promises to form the backbone for the server I/O subsystems of the future. After establishing the motivation for such an architecture, the session discusses in detail the main architectural features, including bandwidth characteristics, configuration options for its topology, and its relationship to the Virtual Interface Architecture, closing by describing plans for incorporating InfiniBand in Windows Server operating systems.

The key success factors for the forecasting task are the established and documented relation between business activity and data processing activity, and the ability to determine the resource consumption demands of new applications. The focus of this session is new application planning, and it defines the relationship between the development process and the capacity planning function.

NT and W2K systems need to be monitored on an on-going basis. This tutorial will cover the issues and procedures for implementing such a process. Key metrics and an approach to establishing on-going measurement and analysis will be detailed.

Most software performance problems are due to fundamental architecture or design problems. Thus they are introduced early in development, but are typically not discovered until late, when they are more difficult and costly to fix. This tutorial presents a systematic, quantitative approach for cost-effectively building performance into software systems. It presents an overview of the software performance modeling techniques, the steps to be conducted during development, performance design patterns for creating systems, and performance antipatterns for recognizing and fixing common problems.