Distributed

Server Load Balancing

Format: Paperback
Author: Tony Bourke
ReleaseDate: August, 2001
Publisher: O'Reilly Media
Rating:

At the title I was impressed
I looked to this book to guide me through numerous pieces and choices of server load balancing on a fair but balanced agenda. But after digging into the content I was forced to re-evaluate my first impression. What I found was far too much information that was simply digested and reformatted from vendor and free areas.

I look to an expert on a topic to give new insight and twists to anything they write. The lack of any personal commitment or recommendation was surprising. Which made the book more into a mid-level technical guide with no path or strategy summary as I would expect. A good beginner type book for the basic information. It was well organized and had decent flow. Chapter 5 seemed to offer the most promise, but never quite concluded with anything concrete.

Needs more examples
It is written for the network administrator, but anyone, including network managers should find it helpful. This book gives an organized but purely descriptive overview of server load balancing and should be helpful to anyone who is approaching the subject for the first time but does not require quantitative assessments of the different methods for doing server load balancing. Network engineers may find the treatment too qualitative, but it could be supplemented by more rigorous material if needed.

Chapter 1 is a brief overview of the history of load balancing and why it is needed in Internet environments. DNS-based, firewall, and global server load balancing are briefly discussed, along with clustering. This is followed in chapter 2 by a brief review of the OSI model and the different components involved in server load balancing. The author cautions that the use of VRRP (Virtual Router Redundancy Protocol) may cause problems if the load balancers become isolated from each other, but does not give any historical or test examples of this. The same holds true for his caution on the use of fail-over cables.

The discussion becomes a little more detailed in chapter 3, wherein the author discusses the actual functioning of a server load balancer. Direct server return is discussed, with its use of MAC address translation and loopback interfaces. Both server-based and switch-based load balancers are briefly discussed, but the author does not want to commit to which is the better architecture. The inclusion of some benchmarking studies would be helpful here.

Performance metrics, such as connections per second, total concurrent connections, and throughput are discussed in chapter 4. Although the discussion is purely descriptive, the author does give a "metrics matrix" that outlines what metrics are important for different types of traffic.

The author gets down to describing the network infrastructure needed to do server load balancing in chapter 5. The different possibilities for network infrastructure are classified according to the IP configuration (flat-based or NAT-based), the return path (bridge path, route path, or DSR), and physical connectivity (one-armed or two-armed). The advantages of the different types of SLB architectures are discussed briefly, but no real test cases or benchmarking studies are included.

The details of flat-based SLB network architectures are discussed in chapter 6. The simplicity of this architecture is emphasized, along with its ability to pass FTP and stream traffic more readily (no real examples given however).

I did not read the rest of the book which covers administrative matters, and so its review will be omitted. .

Disapointing Altogether
. this is the first O'reilly book that I've been disapointed with! Too many purposeless examples and some configs can be downloaded from vendors web site.. am also wondering about the many PETA references for a load balancing book (just off topic).