Customer Reviews

Configuration and Capacity Planning for Solaris Servers

5 star:		(5)
4 star:		(5)
3 star:		(2)
2 star:		(3)
1 star:		(0)

 

 

Every system administrator or reseller consultant configuring systems should read this book to get an understand of the fundamentals of good system design. Much of the theory behind this is common sense, but never seems to occur to people. Brian Wong opens many eyes with this book. I highly recommend it.

The coverage of RAID techniques and their performance characteristics applies as well to HP-UX and WindowsNT as it does to Solaris. The hardware and bus technology gives some insight on how system architecture affects system performance. NFS information applies to any other filesharing service (eg., Coda, IPX, SMB, FTAM, etc.).

The layout is intuitive but limiting. Dividing the types of servers into NFS, Timeshare, DBMS and Internet is helpful. The configuration guidelines under each of these spreads a great deal of useful information throughout the book. This isn't always logical as you are talking about backup policy under DBMS configuration guidelines before covering the backup configuration. Some technologies like PrestoServe are only discussed with NFS when they can be beneficial to other servers (eg., OLTP). The general layout can be improved. A stronger layout would cover all the technologies separately (as done with Storage, RAID and Backup) with details of their configuration benefits. Under each of the server types the technologies that benefit the utilization characteristics could then be mentioned.

An up-to-date edition would be nice, as much of the Hardware used today, isn't even mentioned. PCI is overtaking Sbus, as Sbus overtook VME. FC/AL at 100MB/s is standard. The Cray CS6400 is now Sun's flagship E10k. There are new software features in Solaris, Veritas and Oracle that allow system designers to look at their job from a different perspective. Storage is become more and more centralized with the A7000, IBM ESS, Symetrix and SANS. Solid-state drives are common again. In all, enough technology has changed that a 2nd edition would be well timed.

This book is ok for the older Sun Sparc hardware but its outdated by at least 4 years! There is no coverage of Sun Enterprise servers like the E10000, E3500 or E4000!!!! That makes the book a total waste of money!!! better off, go to docs.sun.com at least its free

The references to the Solaris OS and SPARC machines are of course out of date. Even so, Wong's methodology, analysis of performance factors, and depth of treatment are outstanding. As an Sun-certified educator, consultant, and field engineer, I use this book all the time. I have read it through six or seven times by now, and I still find (or recover) valuable information packed away in the corners. It's indispensable as a methodology reference for scoping systems, although it really should be read cover to cover at least once. Wong's breadth of coverage is absolutely impressive. When I want to show other techs a strategic way to approach system planning, I hand them a copy of this book, first thing. Read this book if you have anything to do with installing or maintaining Sun systems; then ask Brian Wong to please write the update!

The book covers a broad range of generic hardware specific performance/capacity issues well, but lacks information on newer systems (i.e. x500, 6500, 10000), and fails to cover Fibre Channel in enough detail (differences in FCP/SCSI, IP, NDMP, NTP, are not even mentioned).

It goes into great detail about the smallest parameters for SCSI (and other types of disks), RAID architectures, memory and bus timing and much more. Learn why RAID may not be the best means of getting faster disk access. Figure out when you *shouldn't* upgrade that old server in favor of that shiny new one. Most explanations and suggestions are backed up with plenty of real-world data and graphs. It's goal is to help administrators get the most from what they've got and how to scale and appropriately purchase what they want. A very thorough and technical book, and it's rather current (Ultra Enterprise servers are covered throughout). In addition, it details (much) older systems and their architectures. Highly recommended reading for database, system, and web server admins.

Brian Wong has written an extremely useful book.
In the first part he shows techniques for analysing the more common types of servers that exist in the UNIX enviroment. Among the types are the NFS server, the Web server and the RDBMS server. Although Mr. Wong is brisk in his treatment of this material, he sucessfully imparts the flavor of the "back of the envelope" calculations that are most useful in the real world.

The second part of the book is a deeper look at the issues associated with Sun bus architectures, disk subsystems including a good discussion of RAID, backup subsystems, and aspects of the Solaris kernel. Much of this material can be generalized to other vendors by relating published benchmark results of these vendors to the comparable published Sun benchmark results. The Solaris Kernel material was valuable in it's discussion of UFS file system which are widely found in most SVR4 variants.

All in all this is a book that deserves a spot on your bookshelf next to the more formal mathematical treatments of this subject. You'll find yourself reaching for it when you need to write up some quick and dirty analysis to justify a purchase request or to provide a quick estimate of what kind of hardware you'll need to support the next big project your company comes up with.

Some of the basics are still relevant but this book could use some updating since 1997. Also more information on how to figure out how the servers are performing would be nice.

Yes, like the other reviewers I'd like to see a new edition where the examples are based more on things like 36 GB 10k rpm disks instead of 9 GB 7200 rpm disks. The chapters on how the SCSI bus really works, how a disk stores data, what really happens when you use NFS, and how raid works were eye openers for me. Sure some of the topics are a bit out of date but this is a great starting point on your path towards total performance enlightenment.

Great book!! Clear, easy and acurated text!! Great Methodology!! Simple and objective!!

A must have for anyone who needs manage and design unix data centers.

The only problem: This book needs a 2002 update release covering the newest Sun hardware and the most server intensive needs made by web-centric computing.

There is no other book like this, and it does an amazing--MONUMENTAL!!!--job of collecting a lot of the most relevant information together.

My biggest complaint is that this book is NOT a good reference. There is no quick way to find out what is the best way to configure, say, a RAID-5 partition for sequential access for a database with read size of 2KB. You have to literally read through about 100 pages of RAID, disk, SCSI, and filesystem parameter discussions to "compile" the answer. The information that is in the book can be confusing, even conflicting. Methodolgy is often missing or vague.

Possibly, the book suffers from being the first edition. I sincerely hope that a new one will be coming out soon, and will include structural changes as well as information about SUNs latest hardware (the book stops with E6000).

Once again, I do recommend this book to everyone. If you want to read the best book there is on capacity planning in the SUN world, this is the one for you. Just don't expect to get your answers by looking in the index. :)