Solaris 7 Performance Administration Tools / Edition 2

by Frank H Cervone

Overview

-- What it is -- The Solaris 7 product line is Sun's premier software environment for network computing. A UNIX -based operating environment developed by Sun Microsystems. Solaris runs on many workstations from other vendors including Intel. Solaris 7 features a 64-bit environment, mainframe-class reliability features, enhanced PC interoperability, easier installation, administration and configuration.
-- Powerful and flexible -- Solaris 7 is one of the most popular brands of UNIX due to its power and flexibility, but it is difficult to install and set up. However, it is one of the most powerful and stable network operating systems around. Very inexpensive, as well.
-- Widespread with developers -- More than 100,000 copies of Solaris 7 were distributed for Free in the last 9 months to allow developers to work with the and build applications (Source: BusinessWire, 8/5/99).

Product Details

ISBN-13:	9780072122114
Publisher:	McGraw-Hill Companies, The
Publication date:	12/01/2000
Series:	McGraw-Hill Tools Series
Edition description:	BK&CD ROM
Pages:	436
Product dimensions:	6.06(w) x 9.00(h) x 1.15(d)

Read an Excerpt

Chapter 1: Introduction to Performance

1.1 What Is Performance Tuning?

Since the beginning of time, people have been observing the operations of activities, both human and mechanical. While some people engage in these observational activities simply because they enjoy the observational process, most people engage in observational activities specifically to try to predict what will happen the next time an activity is performed. This is no different when the activities are related to a computer or a computing system. It has been said that the first performance check procedure for a computer was introduced after only two computer programs had been written. This check allowed the computer operator to know how far a program had progressed in the calculations it was performing. Considering the instability of the hardware in those early days, this information was invaluable, because it let the operator know where execution could be restarted if there was a failure of some sort.

In today's complicated computer systems, performance observation and measurement are vital. Performance is an essential criterion in the design and operation of a computer system. But what is performance tuning?

For the purposes of this book perforinance tuning is the process of observi ng the operation of an entire computing system and then adjusting the various components of the system based on those observations. The result is that the overall system is more efficient. This definition differs somewhat from more traditional views in that we are not saying we are trying to make a particular component (or indeed even all components) more efficient-we are makingthe entire system moreefficient. In general, this means all components will become more efficient, but this is not necessarily a given. An example will clarify this point.

Consider the case where a system is already in a bad state. Access to disk drives is slow and memory is significantly overcommitted. When memory is overcommitted, the system pages quite a bit and does so very slowly because of problems with the disk drives. Will increasing the speed or number of the disk drives help? Probably, but possibly not. Increasing the speed at which the system is able to page might automatically increase the amount of multitasking. This, in turn, brings the paging rate to a new, higher level that makes the system again come to its knees. So, though there are more processes running and the performance of the disk drive subsystem has been improved, because of the increase in paging, no more real work is being accomplished in a given time quantum. Consequently, the performance of the system has not really been improved.

The above scenario is extremely undesirable. After spending a great deal of money on improving the disk system, one would expect a quantifiable increase in overall system performance. The result of performance tuning should give the system administrator just that: a quantifiable and demonstrable increase in overall system performance.

1.2 What Is Capacity Planning?

Capacity planning complements performance tuning. Whereas performance tuning alleviates some type of bottleneck or slowdown in the system, capacity planning is an ongoing process that ensures adequate computing capacity for an enterprise. By definition then, performance tuning is one of the components of capacity planning.

The desired result of capacity planning is a situation where the users of the system never experience degraded service. Of course, this is a theoretical goal and under most circumstances it is not possible to guarantee adequate service at all times under all situanons. The capacity planner, however, strives toward this goal. If adequate service can be guaranteed to the majority of users most of the time, then one can usually say that the capacity plan is successful.

Of course, there are situations where reliability and availability must be 100 percent, such as in air-traffic control or life-support systems. For most business and academic use, 100 percent availability is a desirable goal as opposed to an absolute requirement

A primary component of the capacity planning process is regular reporting of the status of the system and its capacity. Included As an example, a user from accounting reports to the performance analyst that with the introduction of a new closing program, it now consistently takes ten minutes to close the books for the day. Before the new program, closing the books took only a minute or two. The user has concluded that the new closing program has some type of coding error What the user did not, and cannot, tell the performance analyst, is that a few days before the new closing problem was implemented, an entirely new customer reporting system was introduced. The marketing department is now running several computationally intensive reports at the same time the accounting department is trying to close. Quite possibly, there is absolutely nothing wrong with the new closing program; the problem is that the system load is now much greater. Although the closing program should be investigated for potential problems, the performance analyst should also look at the customer reporting system.

Of course, with a good capacity planning program in place, the above never would have happened to begin with.

1.3.1 Accounting Data Reduction Programs Accounting data, in the context of this book, is information collected regarding the use of computing resources. In the Solaris environment, this information is collected and reported on by various programs in the /usr/lib/acct subdirectory.

Accounting data reduction programs trace their beginnings to the check flags that were inserted into programs in the earlier days of computing. Eventually, these check flags evolved into trace rou tines that allowed the programmer to record the execution of individual instructions as a program ran. By inserting these trace routines into programs at appropriate points, it is possible to gather extensive information on how a program is executing and what factors are influencing its execution This method, however, places additional demands on the computing system and therefore has the potential of significantly influencing the resultant accounting data...

Table of Contents

Section I: Preliminaries.
Introduction to Performance Tuning and Capacity Planning.
Computer Component Overview.
Defining Performance and Selecting Metrics.
Queuing Models.
Section II: Operating System Structure.
Kernel Architecture.
Memory.
Process and Thread Management.
File System Fundamentals.
Networking and Communication Fundamentals.
Section III: System Performance Tuning.
Monitoring Activity and Workbook.
Memory Management.
I/O Performance Management.
Network Performance Management.
Network Services Performance Management.
Kernel Tuning.
Section IV: Capacity Planning.
System Activity.
System Accounting.
Capacity Planning Management.