The information in this publication is subject to change without notice.
COMPAQ COMPUTER CORPORATION SHALL NOT BE LIABLE FOR TECHNICAL OR
EDITORIAL ERRORS OR OMISSIONS CONTAINED HEREIN, NOR FOR INCIDE NT AL OR
CONSEQUENTIAL DAMAGES RESULTING FROM THE FURNISHING, PERFORMANCE,
OR USE OF THIS MATERIAL.
This publication contains information protected by copyright. Except for internal use
distribution, no part of this publication may be photocopied or reproduced in any form without
prior written consent from Compaq Computer Corporation.
This publication does not constitute an endorsement of the product or products that were tested.
The configuration or configurations tested or described may or may not be the only available
solution. This test is not a determination of product quality or correctness, nor does it ensure
compliance with any federal, state or local requirements. Compaq does not warrant products
other than its own strictly as stated in COMPAQ product warr a nties.
Product names mentioned herein may be trademarks and/or registered trademarks of their
respective companies.
1997 Compaq Computer Corporation, Al Rights ReservedDoc No 541A/0697
Introduction
The purpose of this document is to share the knowledge acquired by Compaq Systems Engineers
in the area of configuration and performance tuning of Sybase SQL Server 11 on UnixWare 2.1
on the Compaq Proliant family of servers. It is our desire to deliver the best technical
information possible on a specific topic in a timely manner and in a highly useable format. Any
comments, suggestions and feedback are always appreciated.
The information presented in this document is based on Sybase SQL Server 11 for UnixWare 2.1
and is a result of numerous performance tests executed within the context of an industry-standard
TPC-C benchmark, intern al benchmarking for hardware development and optimization, and
analyzing customer-reported expectations, performance trends, and solutions.
Compaq is an active member of the Transaction Processing Performance Council, and publishes
a number of benchmarks every year proving the superior performance and price-performance
ratios of Compaq servers. The results of these benchmarks can be obtained directly from
Compaq Computer Corporation or from the Transaction Processing Performance Council. The
contact information is listed below.
Even though most of the testing that provided basis for this document was done in the area of
online transaction processing, much of the information presented does apply to other
environments, such as decision support and batch processing. We recommend that you always
experiment before applying any changes to your production server.
Other publications covering these and related topics are listed below:
•Configuring Compaq RAID Technology for Database Servers, Compaq TechNote, P/N
184206-001
• Sybase SQL Server System Administration Guide
• Sybase SQL Server Performan ce and Tun i ng Guide
• Compaq T ech C omm u niqué ‘COMPAQ INSI GHT Ser ver Ma nagemen t’
Transaction Processing Performance Council
c/o Shanley Public Relations
777 North First Str eet, Suite 6000
San Jose, CA 95112-6311
http://www.tpc.org
1997 Compaq Computer Corporati, All Rights ReservedDoc No 541A/0697
Page 2
Configuration and Tuning of Sybase SQL Server 11 for SCO UnixWare 2.1 on Compaq Servers
Tuning Goals
In order to achieve the best performing system possible there are several factors which must be
reviewed. These include optimization of the hardware, the Sybase SQL Server, the operating
system and the application software. This paper will focus on the hardware, Sybase SQL Server
and the OS. It is also important to tune the Sybase application to take advantage of the system.
Due to the diversity of database applications, they are beyond the scope of this paper. This paper
will discuss CPU scalability and tuning, disk controller optimization and I/O tuning, memory
tuning, and network tuning. Also, specific Sybase and UnixWare configuration and tuning
issues will be presented throughout the paper.
System Processor Planning
This section is provided to demonstrate the scalability of various processor configurations on the
Compaq Proliant family of servers and to provide you with some performance information
necessary to determine the best configuration for your environment.
CPU Scalability
The performance information presented in this section was gathered on three Compaq Proliant
models, Proliant 4500 and Proliant 5000 and Proliant 6000, with different system processor
configurations and different amounts of memory. All configurations had one aspect in common:
all were CPU bound.
The tests run were all disk-intensive tests. Disk-intensive tests are designed to access th e entire
range of a database that is many times larger than the data cache of the server. The net effect is
that only a very small portion of the database can fit into the data cache at any point in time, and
a large amount of physical I/O is generated in addition to the transaction log I/O. Such an
environment tests heavily the CPU, memory, and the disk subsystem.
The tests generated a large number of update-intensive transactions and utilized heavily the
transaction log. No hardware or software options were disabled to sacrifice data integrity of the
system.
Figure 1 shows scalability on the Proliant 4500 with Pentium
processors, using 1GB RAM. Figure 2 shows scalability on the Proliant 5000 with
PentiumPro
PentiumPro
6000 with PentiumPro
/166MHz processors, using 1GB RAM and 2GB RAM in the server. The
/166MHz has a 512KB level 2 cache. Figure 3 shows the scalability on the Proliant
200MHz processors with 512K level 2 cache and 4GB RAM. Figure 4
shows memory scalability on the Proliant 6000 with 4 PentiumPro
/100MHz and Pentium/133MHz
200MHz processors with
512K level 2 cache.
1997 Compaq Computer Corporation, All Rights ReservedDoc No 541A/0697
Configuration and Tuning of Sybase SQL Server 11 for SCO UnixWare 2.1 on Compaq Servers
Figure 1: CPU Scalability of the Proliant 4500
CPU Scalabilty on Proliant 4500
3
2.5
2
Page 3
1.5
1
Transaction Rate Normalized
0.5
0
124
Number of CPUs
Figure 2: CPU Scalability for Proliant 5000
CPU Scalabilty on Proliant 5000 with 1GB vs. 2GB RAM
3
2.5
2
1.5
100MHz Pe nt ium
133MHz Pentium w/ 512K
cache
133MHz Pentium w/ 2M
cache
1
Transaction Rate Normalized
0.5
0
1234
Number of CPUs
166MH z PentiumPro w / 1GB
RAM
166MH z Pentium Pro w/ 2GB
RAM
1997 Compaq Computer Corporation, All Rights ReservedDoc No 541A/0697
Page 4
Configuration and Tuning of Sybase SQL Server 11 for SCO UnixWare 2.1 on Compaq Servers
Figure 3: CPU Scalability for Proliant 6000
CPU Scalabi lity for Proliant 6000
3
2.5
2
1.5
1
Transaction Rate Normalized
0.5
0
1234
Number of CPUs
Figure 4: Memory Scalability for Proliant 6000
Memory Scalability for ProLiant 6000
1.6
1.4
1.2
1
0.8
0.6
0.4
Transaction Rate Normalized
0.2
0
1GB2GB3GB4GB
Memory
1997 Compaq Computer Corporation, All Rights ReservedDoc No 541A/0697
Configuration and Tuning of Sybase SQL Server 11 for SCO UnixWare 2.1 on Compaq Servers
Initial CPU Re c ommendations
The choice of the rig ht system p rocessor d ep ends on your envir onmen t. As t echnol ogy evolves,
more powerful processors are becoming available, pushing performance to new levels at very
competitive costs per processing unit. PentiumPro
application servers, especially the PentiumPro
-based system s have become the sta ndard for
166MHz and 200MHz processors.
Our recommendation is to carefully evaluate your environment, and experiment with various
processor configurations, if possible. Always have future growth and expansion in mind. You
may want to start with Pentium
processors for smaller departmental systems, and PentiumPro
processors for higher demand systems. If your environment has many concurrent users, you
should evaluate benefits of multiprocessing.
Befor e upgr a ding the system processor(s), you sh ould closely monitor performance of the system
and tune it from the software perspective. If the performance bottleneck is in software, har dware
upgrades can only par tially improve per formance. It may be more cost effective to tune the
software rather than to purchase hardware upgrades. If the performance bottleneck is clearly at
the system processor(s), upgr a ding to a higher speed processor, or adding another processor can
dramatically improve performance.
CPU Tuning
To monitor CPU performan ce and determine whether your system is CPU bound or not, you can
run the UnixWare per formance monitor, rtpm. This should be run during a normal workload to
get an accur ate reading. You also may use the sar utility to save the information to a file. A
system that is tuned well will have the following CPU char acteristics:
Page 5
• Most of the CPU utilization is in user mode (%usr). Again, this is verified by runnin g
or
and looking at the percentage of CPU time spent in system and in user time. System
rtpm
sar
time (%sys) can be thought of as operatin g system overhead such as time spent in the I/O
subsystem or in system calls. The higher the percentage of user to system time th at you
have, the better. For a system performing mostly Sybase processes, 83 %usr and 17 %sys
is a very good split, assuming there is no CPU idle time. The key is that Sybase should be
getting most of the CPU time. Here is some sample output from
The above exam ple shows a system that is CPU bou nd if all other areas had already been tuned
and optimized for performance. CPU bound means that in spite of your efforts to tune the
system, you cannot get more per formance from it because there is no mor e processin g power left
on the CPU(s). If your database and application are well tuned, and if there is no idle time an d
1997 Compaq Computer Corporation, All Rights ReservedDoc No 541A/0697
Loading...
+ 18 hidden pages
You need points to download manuals.
1 point = 1 manual.
You can buy points or you can get point for every manual you upload.