Section Image

Pub Talk

"High-Volume Web Site Performance Simulator For WebSphere"


By Mr. Noshir C. Wadia
Senior Technical Staff Member
High-Volume Web Sites, Software Group
IBM Corporation

Open to the Public. No fee. 5:00 - 7:00 p.m.
The Pub Talk begins at 6:00 p.m.

As eBusiness and its related requirements grow at "Web speed," a critical issue is whether the IT infrastructure supporting the Web sites has what it needs to provide available, scalable, fast, and efficient access to the company's information, products, and services. More than ever, CIOs and their teams struggle with the challenges to minimize downtime and bottlenecks and maximize the use of the hardware and software that comprises their eBusiness infrastructure. Configuration planning for large Web servers in this environment is becoming an increasingly difficult task. The hardware and software structure of large Web sites is becoming complex and the behavior characteristics of the workloads are at best poorly understood, or, at worst essentially unknown because the workload has yet to be implemented.

The infrastructure supporting most high-volume Web sites (HVWSs) typically has multiple components, which include clients, the network, and multiple layers of machines within the Web server. These multiple machine layers are frequently called tiers with each tier handling a particular set of functions, such as serving content, providing integration business environments, or processing data base transactions. While actual customer implementations can vary widely, the simulator uses the following generalized view of the infrastructure options.

IBM Patterns for eBusiness & Workloads: These are built in workloads for the major business patterns: shopping, trading, and banking. The user can use these pre-built workloads or modify them based on customer's requirements. Users can also define a very specific workload if pre-built workloads do not fit customer environments. We have also developed special algorithms to handle bursty web traffic, which can occur during holiday season or stock market storm. User can define the degree of burstiness in the terms of burst/peak ratio. The user can set a specific performance objective for items such as (1) user visit rate, (2) resource utilization, (3) response time, (4) # of concurrent users, and (5) page view rate. The model will estimate the performance and recommend configurations.

Noshir C. Wadia is a Senior Technical Staff Member in the High-Volume Web Sites organization of IBM's Software Group. He is currently leading a team to develop simulator for performance/capacity estimates. He is also working with large IBM customers on web site performance issues. Mr. Wadia has held many leadership positions in architecture/performance throughout his IBM career developing performance/architecture models for parallel and distributed systems. Mr. Wadia has been author of many papers in the area of performance. He has also been adjunct professor for ten years at State University of New York where he taught Computer Systems Architecture and Design. Mr. Wadia holds a MS in Computer Science from Syracuse University, a MS from Applied Statistics from Purdue University, and a MS in Mathematical Statistics from University of Baroda (India).