May 1996
...........
..........
........
.........
...............
Prepared By
Compaq Systems
Division
Compaq Computer
Corporation
ONTENTS
C
Te st M e thodol ogy and
Te st C onfi gurations
Network Conf igurations
Hardware Configurations
Software Configurations
Performance Charts
C
OMMUNIQUÉ
.
.
.
.
.
Web Server Performance Test Results on
.
.
.
.
.
.
.
Compaq ProLiant 5000 Servers
.
.
.
.
.
This document summarizes the tests performed by Compaq Engineers to measure the
.
.
.
.
overall throughput of the Compaq ProLiant 5000 server running Microsoft Internet
.
.
.
Information Server 1.01 on Microsoft Windows NT Server 3.51 and processing client
.
.
.
requests for HTML pages as well as pages containing CGI. The Compaq ProLiant 5000
.
.
.
scales better than the ProLiant 4500 under moderate to heavy request loads. The test
.
.
.
data shows a 49% improvement on the ProLiant 4500 and a 56% improvement on the
.
.
3
.
ProLiant 5000.
.
.
4
.
.
5
.
In the CGI testing that we outline below, we saw a 53% increase with the ProLiant 5000
.
.
5
.
single processor over an equivalently configured ProLiant 4500. In the dual-processor
.
.
6
.
configuration, we saw a 60% increase in performance with the ProLiant 5000 when
.
.
.
.
compared to the equivalent ProLiant 4500 configuration. In the Webstone compatible
.
.
.
tests that were run, the performance increase for a dual-processor ProLiant 4500 versus
.
.
.
a dual-processor ProLiant 5000 is approximately 20%. This smaller level of increase is
.
.
.
due to the nature of the Webstone load which is not processor intensive. However, in a
.
.
.
comparison of a single processor ProLiant 4500 to the single processor ProLiant 5000,
.
.
.
there was an increase realized of 68%. This increase is attributed to the fact that the
.
.
.
.
Webstone load is causing enough network traffic to cause the single ProLiant 4500
.
.
.
processor to be nearly saturated where the faster ProLiant 5000 was able to handle the
.
.
.
test load with greater ease.
.
.
.
.
.
The testing demonstrates that the ProLiant 5000 outperforms the ProLiant 4500 in
.
.
.
various proportion client requests (light request loads, medium request loads, and heavy
.
.
.
request loads). In summary, the test results demonstrate that Compaq ProLiant servers
.
.
.
are an excellent web server platform.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
OMMUNIQUÉ
C
.
.
.
OTICE
N
.
.
.
.
.
The information in this publication is subject to change without notice.
.
.
.
.
.
.
.
C
OMPAQ COMPUTER CORPORATION SHALL NOT BE LIABLE FOR TECHNICAL
.
.
.
OR EDITORIAL ERRORS OR OMISSIONS CONTAINED HEREIN
.
.
INCIDENTAL OR CONSEQUENTIAL DAMAGES RESULTING FROM THE
.
.
.
FURNISHING
.
.
.
.
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 warranties.
.
.
.
.
Product names mentioned herein may be trademarks and/or registered trademarks of their
.
.
.
respective companies.
.
.
.
.
.
Compaq, ProLiant, and NetFlex, registered United States Patent and Trademark Office.
.
.
.
.
Netelligent is a trademark of Compaq Computer Corporation.
.
.
.
.
.
Other product names mentioned herein may be trademarks and/or registered trademarks of their
.
.
.
respective companies.
.
.
.
.
©1996 Compaq Computer Corporation. All rights reserved. Printed in the U.S.A.
.
.
.
.
Microsoft, Windows, Windows NT, and Windows NT Server are trademarks and/or registered
.
.
.
trademarks of Microsoft Corporation.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Web Server Performance Test Results on
.
.
.
.
Compaq ProLiant 5000 Servers
.
.
.
First Edition (May 1996)
.
.
.
Document Number 210A/0596
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2
(cont.)
,
PERFORMANCE, OR USE OF THIS MATERIAL
,
NOR FOR
.
OMMUNIQUÉ
C
.
.
.
.
.
.
T
EST METHODOLOGY AND TEST CONFIGURATIONS
.
.
.
.
.
The tests were first conducted on a 10-megabit network using eight clients running multiple
.
.
.
threads to simulate up to a maximum of 256 threads. Compaq Engineers encountered a network
.
.
.
bottleneck at light to moderate request loads on the 10-megabit network. Because of this, the tests
.
.
.
were run over a 100-megabit network to ensure that the network was not the bottleneck. The
.
.
server and clients were connected to the same 100BaseTX LAN using Compaq 100BaseTX NICs.
.
.
.
The test used Microsoft’s WebCat analysis tool with the Webstone file set to measure the server
.
.
.
throughput in bytes per second. All tests were run using the same request load.
.
.
.
.
.
Performance analysis on web servers heretofore are an unknown. There has been some work done
.
.
.
in this area, but to date little has come about for an industry standard tool. Some companies have
.
.
.
adopted a metric and tool from Silicon Graphics Inc. (SGI) called Webstone. This tool uses the
.
.
.
standard Hyper Text Transfer Protocol 1.0 (HTTP) method GET to request documents from the
.
.
.
web server under test. These requests can be generated from one or many clients to artificially
.
.
simulate a load to any web server.
.
.
.
.
.
Web server performance is relative. The information you base performance characteristics on will
.
.
.
depend on the type of load expected for the web server under test. Because of this, Compaq
.
.
.
determined that the Webstone file set should be used in our tests, since it is the de facto standard
.
.
.
at this time. The WebCat load generation tool supports the Webstone file set, which enables us to
.
.
.
use this tool to generate a load that approximates the results one would get using Webstone.
.
.
.
.
.
EBSTONE FILE DISTRIBUTION
W
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
Percentage of
time requested
40% 2K, 3K
25% 1K, 5K
15% 4K, 6K
5% 7K
4% 8K, 9K, 10K, 11K
4% 12K, 14K, 15K, 17K, 18K
6% 33K
1% 200K
(cont.)
Random file
size/selection
OMMUNIQUÉ
C
.
.
.
.
While Webstone is useful in comparing performance across platforms, actual customer use will
.
.
.
include requests for pages containing both HTML and CGI content. To highlight the processing
.
.
power of the ProLiant 5000, the Common Gateway Interface (CGI) file set was selected as an
.
.
.
alternate mechanism to test the server’s ability to process requests. CGI is a term that is used for
.
.
.
web servers to denote an application that requires processing and is requested by a web client but
.
.
.
executes on the web server. The CGI test we used here distributes the load between CGI and text
.
.
.
files according to the following table:
.
.
.
.
.
.
CGI/HTML
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
Percentage of
time requested
2% 256K
2% 512 Bytes
2% 1K
3% 2K
3% 3K
2% 4K
3% 6K
1% 8K
3% 16K
2% 32K
1% 64K
1% 256K
75% CGI Test
Network Configurations
Web Client Controller
(cont.)
FILE DISTRIBUTION
Random file
size/selection
WebServer
4500 / 5000
100 Mbps, Dedicated LAN
Client 1 Client 8
OMMUNIQUÉ
C
.
.
.
.
.
.
Hardware Configurations
.
.
.
.
.
.
Server Hardware Configuration: ProLiant 5000, 2-Pentium Pro/166 Mhz Processors, 64 MB
.
.
.
RAM, SMART-2 PCI Controller, Netelligent 10/100 PCI Ethernet Network Interface Controller,
.
.
One 2.1 GB Fast-Wide-SCSI-2 Drive.
.
.
.
.
.
.
Server Hardware Configuration: ProLiant 4500, 2-Pentium/133 Mhz Processors, 64 MB RAM,
.
.
.
SMART-2 PCI Controller, 100BaseTX NetFlex-3 Ethernet Network Interface Controller, One 2.1
.
.
.
GB Fast-Wide-SCSI-2 Drive.
.
.
.
.
.
.
Client Hardware Configuration: ProLiant 2000, 2-P5/90 Mhz Processors, 32 MB RAM,
.
.
.
SMART EISA Controller, 100BaseTX NetFlex-3 Network Interface Controller, One 2.1 GB Fast-
.
.
.
SCSI-2 Drive.
.
.
.
.
.
.
.
.
.
.
.
.
Software Configurations
.
.
.
.
.
.
Server Software Configuration: Windows NT Server 3.51, Service Pack 4, Microsoft Internet
.
.
.
Information
.
.
.
.
.
.
.
.
.
Client Software Configuration: Windows NT 3.51, Service Pack 4.
.
.
.
.
.
.
Test Tool: Microsoft WebCat (Web Server Analysis Tool)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
(cont.)
Server Version 1.01, Compaq SSD Version 1.18.
OMMUNIQUÉ
C
.
.
.
.
.
.
Performance Charts
.
.
.
.
.
.
The two charts below show the performance of the ProLiant 5000 in comparison to the ProLiant
.
.
.
4500. The charts show the scalability of the ProLiant 5000 and the Pentium Pro processors. The
.
.
CGI Performance chart shows a drastic improvement in comparison to the Webstone Performance
.
.
.
chart, since CGI requests are very processor intensive.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6
45
40
35
30
25
20
Mbs / Sec
15
10
5
0
Figure 1. ProLiant 5000 vs ProLiant 4500 Webstone Performance
(cont.)
Webstone Performa nce
5000, 2-processor, 64-MB
5000, 1-processor, 64-MB
4500, 2-processor, 64-MB
4500, 1-processor, 64-MB
16 32 64 128 256
Number of Clients
OMMUNIQUÉ
C
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
7
6
5
4
3
Mbs / Sec
2
1
0
Figure 2. ProLiant 5000 vs ProLiant 4500 CGI Performance
(cont.)
CGI Performance
16 32 64 128 256
Number of Users
5000, 2-processor, 64-MB
5000, 1-processor, 64-MB
4500, 2-processor, 64-MB
4500, 1-processor, 64-MB
Loading...