Capacity Planning for Web Performance: Metrics, Models, and Methods [Paperback]

Virgillio Almeida (Author), Virgilio A. F. Almeida (Author)

Book Description

Publication Date: June 1998

This book will discuss the problem of Capacity Planning and Performance Analysis in Web Server, Inranet and Client/Server environments. It will identify problem areas where capacity planning and performance analysis are critical concerns: arrival rate, through-put, response time, service demand, workload, delay, bottleneck, and saturation. It will discuss protocol (HTTP & TCP/IP) and workloads (access to HTML documents, graphics, etc.). It will show how to access existing capacities and how to plan for future capacities. It will discuss benchmarking metrics, global systems problems, workload forecasting, etc.

Editorial Reviews

Amazon.com Review

Given your server hardware and Internet connection, how many users can your Web site handle? Capacity Planning for Web Performance can show you the techniques for estimating and planning effectively for your Web site's workload, both for today and tomorrow. This textbook-style treatment of the topic presents concepts and formulas for making sure your Web infrastructure is up to the task.

In early sections of this book, the authors introduce the basic concepts of capacity planning and some of the Web-specific issues that you must overcome for effective planning. (For instance, Web traffic is "bursty"--as any Webmaster will attest--and can fluctuate greatly.) Short chapters on system architectures, from traditional client servers to today's Web-centered thin clients, are discussed, as are the basics of TCP/IP and HTTP.

Subsequent sections discuss how to measure performance on your system, using tools such as Web benchmarks, and how to provide formulas for estimating how much hardware is required for "acceptable" performance.

Throughout Capacity Planning for Web Performance, there's a fair amount of mathematical detail. Though there are plenty of real-world examples, this is not a guide to just tweaking Web servers; this is a highly technical guide to state-of-the-art thinking on issues of measuring performance on the Internet. Applying metrics to Web performance can benefit any Web-minded business, though it will probably take the technically savvy reader to effectively execute this knowledge. --Richard Dragan

Review

Reader Reviews From: Peter J. Denning Many have said that the Web is too amorphous and chaotic to permit meaningful performance forecasts. Almeida and Menasce demolish this myth. Throughput, response time, and congestion can be measured and predicted, all using familiar tools from queuing networks that you can run on your own computer. There is no other book like this. It is a first. Quote by Leonard Kleinrock, Professor of Computer Science, UCLA "This is a welcome approach to the performance analysis of todays web-based Internet. It is a useful and practical treatment that is eminently accessible to the non-mathematical professional. An impressive feature the authors provide is to deal directly with the fractal nature of web-based traffic; no simple and practical treatment has been offered before, and theirs is a timely contribution." From: Jim Gray, Senior Researcher, Microsoft Research "This excellent book gives a quantitative and pragmatic approach to measuring, analyzing, and understanding web servers. It presents a good tutorial on the performance issues of web servers, and presents the analytic tools needed to model them. Web servers have bursty and highly-skewed load characteristics. This book presents a new way to model, analyze, and plan for these new performance problems. The book is a valuable resource for students and for web-administrators." From: Jeffrey P. Buzen, Chief Scientist and CoFounder BGS Systems "This book takes the mystery out of analyzing Web performance. The authors have skillfully culled through more than "a" DELETE twenty-five years of performance related research, and have selected the results that are most critical to Web performance. They have also developed important new material that deals directly with the special properties of applications that run on the Web. With everything together in a single volume, Menasce and Almeida have created a superb starting point for anyone wishing to explore the world of Web performance".

See all Editorial Reviews

Product Details

• Paperback: 350 pages
• Publisher: Prentice Hall PTR (June 1998)
• Language: English
• ISBN-10: 0136938221
• ISBN-13: 978-0136938224
• Product Dimensions: 9 x 7 x 1.1 inches
• Shipping Weight: 1.6 pounds
• Average Customer Review: 4.6 out of 5 stars  See all reviews (7 customer reviews)
• Amazon Best Sellers Rank: #795,435 in Books (See Top 100 in Books)
  • #19 in Books > Computers & Technology > Web Development > Website Analytics

More About the Author

Discover books, learn about writers, read author blogs, and more.

Customer Reviews

Most Helpful Customer Reviews

24 of 24 people found the following review helpful:

5.0 out of 5 stars Excellent Tutorial and Reference for Web Performance Models, November 20, 1998

By 

Ted Hruzd (New Jersey) - See all my reviews
(REAL NAME)   

This review is from: Capacity Planning for Web Performance: Metrics, Models, and Methods (Paperback)

If you thought Web architectures were too complex for modeling, you are wrong ! This text explains all possible major components of Web transactions - from TCP/IP, http, CGI, proxy and cache servers, browsers, and networks, in detail. It also explains and adapts various utilization, queue, and response time models to performance analysis and capacity projections. This text is outstanding as both a tutorial and reference. Particularly useful are many real world examples with solutions based on the models. The models are available as Excel worksheets. I recommend this text for all who are serious about designing Web applications that scale well and that are responsive to users. --- Ted Hruzd, performance analysis / capacity planning in the securities industry since 1984; platforms: Tandem, HP UX, and NT --- [e-mail]

21 of 21 people found the following review helpful:

5.0 out of 5 stars A must for any serious web site architect., September 16, 1998

By A Customer

Amazon Verified Purchase

This review is from: Capacity Planning for Web Performance: Metrics, Models, and Methods (Paperback)

As a technical consultant in interactive solutions for a large systems integrator, I am confronted on every project with the capacity planning issues related to e-comm applications. The estimation of sizing and performance are often considered "too complex" and lead to high-level deliverables that eventually cannot serve as an acceptable basis for service level agreements in a company. This book, gives an a methodology that leads to metrics that model your system. The approach is practical and the accompanying examples (and excel spreadsheets) are very concrete. As I don't have the time to go back to the gory details of queuing theory, I find the formulas well explained and helpful to derive the required SLAs and other metrics needed to roll out a system where the performance is understood. Really useful!

16 of 16 people found the following review helpful:

4.0 out of 5 stars Good introduction for the beginner, June 11, 2001

By 

Dr. Lee D. Carlson (Baltimore, Maryland USA) - See all my reviews
(VINE VOICE)    (HALL OF FAME REVIEWER)    (REAL NAME)   

This review is from: Capacity Planning for Web Performance: Metrics, Models, and Methods (Paperback)

The modeling of the Internet has become extremely important in recent years as it continues to grow in leaps and bounds. Network architects have to become very aware of the performance issues when they design networks that will be integrated into this elaborate spider of clients, servers, routers, and switches. The issues in the modeling of global networks are extremely complex and involve very advanced mathematical techniques in order to do the job effectively. The authors of this book however have written an introduction to Web modeling that is written at a level appropriate for network designers and the beginning modeling engineer. They employ Excel spreadsheets and C code to assist in the modeling efforts, and these packages are available on an accompanying CD.

After a brief discussion of the issues concerning capacity planning, Web server, Intranet, and ISP performance in Chapter 1, the authors move on to defining and characterizing client/server systems in the next chapter. After a brief overview of the history of the Internet, they discuss LANs and WANs, and a quick treatment of protocols. The TCP protocol is considered in somewhat more detail because of its importance in network performance.

The quantitative analysis of performance in client/server environments is begun in chapter 3, wherein the authors begin with communication-processing delay diagrams to illustrate how requests spend time at each resource. This is done for both a 2-tier and a 3-tier C/S architecture, and the authors detail how disk subsystems contribute to the service time at a disk. An elementary iteration technique is used to compute the disk utilization. A very interesting and detailed discussion of the RAID-5 disk array is given. Some elementary queuing theory is discussed, using the assumption of flow equilibrium. A simplified summary of the utilization, forced flow, service demand, and Little's laws is also given without resorting to complicated mathematics.

Performance issues in Intranets and Web servers are the topic of the next chapter, and most importantly, the authors outline the differences between HTTP 1.0 and HTTP 1.1. The role of the proxy server and its contribution to performance is also discussed, along with Web cluster architectures. The authors first mention the role of burstiness in this chapter, but do not give an in-depth mathematical discussion.

In chapter 5, the authors give a step-by-step methodology for capacity planning for C/S systems. Workload characterization, data collection issues, model validation, and forecasting are all discussed quantitatively with more details in later chapters.

How to characterize the workload quantitatively is the subject of the next chapter, in terms of a business, functional, and resource-oriented methodology. The authors discuss briefly workload models from a non-mathematical point of view, with parametrized models given the emphasis. The calculation of the parameters is given a more detailed and mathematical treatment, with distance measures and clustering algorithms outlined. Self-similarity in network traffic is first mentioned here, but not discussed from a rigorous mathematical perspective. The authors do however give a rudimentary method for calculating the burstiness.

Benchmarking is discussed in Chapter 7, with the authors detailing the most common approaches to this activity, and mention the most cited benchmark sources, including SPEC, TPC, AIM, and NNBB. The authors divide benchmarks into two categories, component-level and system-level, and discuss CPU performance benchmarking, file server performance, and transaction processing systems as examples of these two categories. Web server benchmarking is also discussed in the context of the two most popular benchmarks: Webstone and SPECweb. Webstone uses Little’s Law to derive a metric called Little’s Load Factor, which gives the average number of connections open at the Web server at a particular time during a network test. Their discussion is very helpful for network modelers who need an introduction to the current benchmarks used in network testing and planning.

The authors fortunately get even more mathematical in the next two chapters on system-level and component-level performance models. Various queuing models are analyzed assuming operational equilibrium, which the authors assume for all models in the book, and which means that the number of requests initially is equal to the number at the end of the observation interval. State transition diagrams are introduced, but the mathematical formalism used is not based on one from stochastic processes, but instead is more phenomenological. The authors employ mean value analysis to solve closed queuing networks with the EXCEL spreadsheets nicely illustrating the results.....

The last chapter of the book discusses how to obtain network performance data experimentally. This can be a difficult task, but the authors do a good job of discussing the possible strategies one can use to collect this data, and give a brief overview of the commercially available network monitors available for this purpose. The difficult job of parameter estimation using measurement data is also discussed in some detail. The authors refer to their other book however for a more thorough treatment of validation and calibration techniques.

The authors have written a fine book here, and will serve well the person first beginning in network modeling and the network designer who needs to understand performance issues. After reading this book, and with some more mathematical preparation, readers can then move on to more sophisticated treatments of the mathematical and simulation modeling of networks.