Optimizing Network Performance with Content Switching: Server,...

Features

• Cover Type: Hard Cover with 288 pages
• Published by: Prentice Hall PTR July 12, 2003
• Written in: English
• ISBN 10 Number: 0131014684
• ISBN 13 Number: 978-0131014688
• Book Dimensions: 9.4 x 7.6 x 0.7 inches
• Weighs: 13.6 ounces

Back Cover Copy


Enrich the user experience while keeping control of your network

Networking professionals: At last, comprehensive coverage of content switching technology is in your hands! Optimizing Network Performance with Content Switching: Server, Firewall, and Cache Load Balancing, by topic experts Matthew Syme and Philip Goldie, provides sharp insights into this wide-ranging, rapidly changing technology. You'll learn the concepts, the background, and the architectures of content switching, as well as how it works and where and why it should be deployed to maximize the performance of your networks and servers. Fueled by increasing bandwidth and ever-richer content, this technology is fast becoming a standard in enterprise and ISP networks. Inside you'll find:

• Full explanations of TCP/IP, HTTP, DNS, and other network fundamentals
• Vendor-independent coverage of the key content switching applications--server load balancing, Web cache redirection, application redirection, firewall load balancing, and more!
• Detailed diagrams that illustrate clearly the deployments and topologies for all applications and variations covered in the book
• A running fictional case study that translates theory into practice, with all applications demonstrated in example implementations
• Advanced topics including TCP/IP flows and sessions, delayed binding, URL parsing, cookie persistence, server health checking, and load balancing metrics

The advanced topics in Optimizing Network Performance with Content Switching are tailored for engineers with layer 2 and layer 3 experience looking for a comprehensive explanation of the inner workings of content switching. This book is also ideal for managers, application developers, security administrators, sales people, and other professionals who require a firm understanding of the basics of the technology.

About The Author


MATTHEW SYME is a product specialist for the Alteon portfolio of Nortel Networks. After exposure to content switching in early 2000 he joined Alteon WebSystems. Matthew has over a decade of experience in networking, working on major infrastructure projects in South Africa, United Kingdom and Australia.

PHILIP GOLDIE is a Product Specialist for the Alteon portfolio within Nortel Networks. Previously, he was a systems engineer for Alteon WebSystems, one of the pioneers of content switching. In nearly a decade working in the networking industry, Philip has been a customer, a reseller, and a vendor.

Reader Reviews
The concept of content switching is relatively new, but for those engineers who work with content switches in real networks, their behavior can be very unpredictable and can conflict with existing hardware and applications. Indeed, there are a few instances where content switch server load balancing competes with the load balancing done in some multi-tier environments running certain types of protocols, such as T3, coupled with the use of WebLogic activated over an application cluster. These issues are not discussed in the book, but instead the authors give a purely descriptive and qualitative overview of the history, functioning, and use of content switches. However, from the title, a reader might expect that the book is a rigorous, mathematical study of how the use of content switching can enhance network performance. Such a study is not done in this book, but there is a great need for it for anyone who is considering the use of content switching. As the authors explain it, content switching has its origins in the rise of the Internet and involves using a single point by which a session can be established. A virtual IP address (VIP) is configured on the content switch, which allows a user to connect to the single point, leaving the content switch to set up a dialog with the server or do appropriate load balancing over a collection of servers. After a review of the OSI model, the authors begin a discussion of switching at different layers. The claim that the information in the different layers can be used to perform traffic forwarding decisions which are "intelligent." Their use of the term "intelligent" here may be inappropriate though, as there is nothing about content switching that can be deemed intelligent, at least from the standpoint of this reviewer. Certainly a content switch is doing something a lot more involved than an ordinary "switch", but to be deemed intelligent it must learn from its past, or learn from its mistakes, or take action that it deems necessary for improving the user's experience. The Virtual Router Redundancy Protocol (VRRP) is discussed in the book, due to its importance in content switching. This discussion is helpful, especially for those who do not have knowledge of this protocol, and how it is used to eliminate single points of failure within content switching topologies. VRRP allows the elimination of single points of failure within content switching topologies and allows the grouping of two or more IP addresses so that they appear to surrounding devices as a single logical IP address. The difference between virtual services and application redirection is fundamental to content switching according to the authors. Server load balancing and Web cache redirection are examples illustrating this difference. In server load balancing, the content switch has a virtual IP address to which a client will attach. This is the origin of the designation as a "virtual service", in that the destination address of the client's requests is owned by the content switch. Web cache redirection involves manipulation of the client traffic as it passes through the content switch on its way to its destination. Server load balancing is commonly (and incorrectly) equated to content switching in some organizations. The authors give a fairly detailed and descriptive overview of layer 3 server load balancing, giving conservation of address space and simplicity as the two primary reasons for deploying it. They also mention, but do not discuss in any great detail, the use of dynamic routing protocols, such as OSPF, RIP, and BGP, in content switching. A more detailed discussion, coupled with performance data, would be very helpful to those readers who are considering using content switching in conjunction with these protocols. They do discuss how to implement high availability in server load balancing using multiple content switches and routers, and using multihoming. The different load balancing metrics, such as least connections, round robin, IP address hashing, response time, bandwidth, and server weighting, are also discussed. Unfortunately, the authors do not include any real test cases or benchmarking studies that illustrate more quantitatively the differences between these metrics. They do however point out the use of server agents to gain information on CPU and disk utilization, and memory performance so as to allow the content switch to better influence traffic flow and session distribution. The use of server agents is growing, and represents the bulk of future development, so it would have been helpful if the authors had included a study of the tradeoffs in using these agents in real networks and servers that are deploying content switching. Criteria to designate a server as being "healthy" is also important in server load balancing, and the authors discuss some common approaches to checking server health at layers 2 through 7. These include link-based, ARP, ICMP, TCP, and application health checks. The authors are very aware of the care needed to designate a server as being healthy, but they do not site and case studies that illustrate the real-world behavior of "unhealthy" versus "healthy" servers. Claims are made that the deployment of layer 4 load balancing does increase user response time in a hypothetical "test case" that the authors include in the book, but no explicit data is given that illustrates the situation before and after the load balancing is used. Layer 7 or "content aware" server load balancing is usually what is thought of when content switching is mentioned. The authors devote and entire chapter to it, and claim many advantages for its use. Applications of it include HTTP URL parsing, cookies, and header inspection, FTP and DNS parsing, and RTSP stream parsing, all of which the authors discuss in some detail. The `delayed binding' by the content switch makes the decision-making more complicated than layer 4 load balancing, and the authors discuss in detail the differences between immediate versus delayed binding of sessions. Once again though, no real-world examples are given that would illustrate the issues that arise in layer 7 server load balancing.