HP D7171A - NetServer - LPr Instruction Manual

HP NetServer
Performance and Sizing
Guidelines
Microsoft Windows NT 4.0 Server,
Terminal Server Edition
January 1999

Hewlett-Packard Company 2
Notice
The information contained in this document is subject to change without notice.
Hewlett-Packard makes no warranty of any kind with regard to this material, including, but not
limited to, the implied warranties of merchantability and fitness for a particular purpose.
Hewlett-Packard shall not be liable for errors contained herein or for incidental or consequential
damages in connection with the furnishing, performance, or use of this material.
The use of trademarks or other designations is for reference purposes only. The following
trademarks are used in this manual:
Intel, Pentium and Xeon are registered trademarks of Intel Corporation. Microsoft and MS-
DOS are U.S. registered trademarks of Microsoft Corporation. Windows and Windows NT are
trademarks of Microsoft Corporation. Citrix WinFrame and ICA are registered trademarks of
Citrix Systems, Inc.
Hewlett-Packard
Network Server Division
10955 Tantau Avenue
Cupertino, California 95014
© Copyright 1999, Hewlett-Packard Company

Hewlett-Packard Company 3
Contents
2 TEST SUMMARY............................................................................4
Introduction......................................................................................................................4
Processors.........................................................................................................................5
Memory............................................................................................................................7
Network............................................................................................................................7
Client Solutions ................................................................................................................8
3 SIZING ACTIVITY..........................................................................10
Methodology...................................................................................................................10
Test Configuration .......................................................................................................10
Visual Test Scripts.......................................................................................................14
Test Procedures............................................................................................................16
Test Results ....................................................................................................................16
Elapsed Time...............................................................................................................17
Process Utilization........................................................................................................18
Network Utilization......................................................................................................24
Memory Utilization......................................................................................................26

Hewlett-Packard Company 4
2 Test Summary
Introduction
In a Microsoft® Windows NT® 4.0, Terminal Server (TSE) multiuser
environment users share hardware and software resources on the server.
These resources include processors, memory, and storage as well as operating
system resources and application resources such as configuration data (e.g.,
location of directories and files, arrangement of application within groups, and
desktop appearance).
In order to help standardize sizing information, Microsoft has defined three
categories of users:
• Task-based users. A task-based user typically runs a single application
used for data entry, for example a line-of-business application written in
Visual Basic (POS, Order Entry applications, SAP, and so on).
• Typical user. A typical user runs one or two applications but usually runs
only one at a time. A typical user runs applications whose data processing
demands on the system are not heavy, for example a productivity
application such as a word processor, a web browser or an e-mail client.
• Advanced users. An advanced user is a more sophisticated user who runs
three or more applications, and often has several applications active at the
same time. An advanced user runs applications whose data processing
demands on the system are heavy.
For the sample configurations described in the following table, Microsoft
provides, for each type of user, the recommended number of users that can be
supported (Microsoft Window NT Server, Terminal Server Administrator's
Guide).
Configuration Task-based users Typical users Advanced users
Single processor Intel®
Pentium®Pro 200MHz
128MB RAM
25 15 8
Dual-processor Pentium Pro
200MHz 256MB RAM 50 30 15
Quad-processor Pentium
Pro 200MHz 512MB RAM 100 60 30
Table 1

Hewlett-Packard Company 5
Processors
In the TSE environment, all users contend for the same processors. In this
environment a single CPU-intensive application can adversely effect the
performance of all users on the system.
As expected, our results also show that as the CPUs become saturated,
performance deteriorates significantly and client sessions can become starved.
As a result, solutions consisting of well behaved, predictable applications (e.g.,
order entry applications) are ideally suited for this environment.
The following table contains general guidelines, for the number of users that
can be supported by on a sample set of HP NetServer configurations.
HP NetServer
Configuration Task-based users Typical users Advanced users
LX Pro: Quad processor
Pentium Pro 200 MHz
768MB RAM
100 60 30
LPr: Dual processor
Pentium II 450 MHz 768MB
RAM
110 65 32
LH 3: Dual processor
Pentium II 450 MHz 768MB
RAM
110 65 32
LXr 8000: Quad Processor
Pentium II Xeon™400
MHz 1GB RAM
200 120 60
Table 2
Resource requirements can vary significantly depending on the application mix
and client usage pattern. Therefore, whenever possible, HP recommends
sizing the server using the targeted application environment.

Hewlett-Packard Company 6
Our tests also show that TSE scales well, evenly sharing the processors
among the client sessions.
The following chart shows the average and maximum processor utilization for
each of the four processors during the 1, 12, 24, 36, 48, 60 and 72 client runs
on the HP NetServer LXr 8000. This chart illustrates how evenly processing is
shared among the processors.
% Processor Utilization
0
10
20
30
40
50
60
70
80
90
100
Number of Client Sessions
Processor 0(Avg) 0.01 0. 05 0. 3 1.66 5. 19 11.14 25.01 37.64
Processor 1(Avg) 0.04 0.08 0.37 2.02 4.77 11.81 25.26 37.66
Processor 2(Avg) 0.01 0.04 0.45 1.68 5.18 11.34 25.52 38.05
Processor 3(Avg) 0.1 0. 03 0.59 1.81 5.13 11.63 25.35 37.79
Processor 0(Max) 0.21 0.78 3.59 6.56 27.99 61.94 100 100
Processor 1(Max) 1.2 1.46 4.35 11. 03 27.34 62.4 100 100
Processor 2(Max) 0.42 0.47 5.47 7.15 28.41 63.85 100 100
Processor 3(Max) 1.09 0.52 5.93 7.54 28.12 63.34 100 100
0 1 12 24 36 48 60 72
Figure 1

Hewlett-Packard Company 7
Memory
As with processors, configuring the server with insufficient memory will
seriously impact performance. Since memory requirements will vary, it is
extremely important to accurately size your server configuration using the
targeted application environment.
For Terminal Server, Microsoft requires 32MB of RAM, plus an additional 4 to
8MB RAM for each user to support running the user's applications on the
server (Microsoft Window NT Server, Terminal Server Administrator's Guide).
Memory utilization during our benchmark test averaged between 5 and 9MB
per user.
Network
As with processor and memory utilization, network utilization will vary
depending on user activity. As a general rule of thumb, Microsoft recommends
2 - 6Kbps per user (Microsoft Terminal Server Capacity Planning, White
Paper).
Network activity during our benchmark test averaged between 2 and 5Kbps per
user.

Hewlett-Packard Company 8
Client Solutions
Microsoft Terminal Server supports a full range of client choices including
Windows-based Terminals (WBTs), Palmtop PCs, NetPCs, and PCs. This
flexibility allows customers to select the client technology that best meets their
needs. HP, as a leading provider of both client and server technology, is
uniquely positioned to offer a full range of client/server solutions for the
Microsoft Terminal Server market.
Our tests have shown that the type of client (i.e., WBT, NetPC, or PC) is not a
significant factor when sizing the server. In order to illustrate this result the
same test script was run on both a PC configuration and a WBT configuration.
12
NetSe rverLHx
10 Mbits
Microsoft Terminal Server
HP NetServer
Microsoft Terminal Server Clients
HP Vectra
HP J31218A
PC Configuration
12
NetServerLHx
10 Mbits
Microsoft Terminal Server
HP NetServer
Microsoft Terminal Server Clients
HP Windows-based Terminals
HP J31218A
WBT Configuration
The next three charts illustrate that server behavior (i.e., performance and
resources utilization) does not vary significantly with a change of client
technology.
The following chart shows the average elapsed time recorded during 1, 2 and 3
client runs using the HP Vectras and HP Windows-based Terminals.
Average Elapsed Time
2000
2100
2200
2300
2400
2500
2600
2700
2800
2900
3000
Number of Client Sessions
PCClients 2440 2435 2431
WBTClients 2444 2438 2434
123
Figure 2

Hewlett-Packard Company 9
The following chart shows the average processor utilization recorded during 1,
2 and 3 client runs using the HP Vectras and HP Windows-based Terminals.
Average Processor Utilization
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Numberof Client Sessions
PCClient s 0.12 0.23 0.34
WBTCl ient s 0.13 0.23 0.35
123
Figure 3
The following chart shows the % Network Utilization recorded during 1, 2 and 3
client runs using HP Vectra and HP Windows-based Terminals.
Network Segment - Average % Network
0
0.5
1
1.5
2
2.5
3
N
umber of Client
PC 0.1
0
.26
0
.37
WBT 0.1
0
.26
0
.37
1
2
3
Figure 4

Hewlett-Packard Company 10
3 Sizing Activity
Methodology
This sizing activity consisted of running the same test script on various HP
NetServer configurations and then comparing the results.
A Visual Test 4.0 script, based on the Microsoft Office 95 product suite, was
used to automate and standardize the interaction of users (i.e., clients) with
Microsoft Windows NT 4.0 Server, Terminal Server Edition using the Microsoft
RDP protocol. (Visual Test is a product of Rational Software,
www.rational.com.)
For each TSE client session, the time to execute the test script (i.e., the
elapsed time) was recorded. In addition to elapsed time, additional system
utilization statistics were gathered using the Microsoft Performance Monitor.
Test Configuration
The test configuration consisted of an HP NetServer connected to 1 - 12 HP
Vectra PCs. PCs, rather than WBTs, were used in the test environment
because the PCs allowed multiple TSE client sessions to be simulated using a
single client system.
To create the network, two sets of six clients were each connected to cascaded
HP J3128A AdvanceStack 10Base-T hubs.
12
NetServerLHx
10 Mbits
Microsoft Terminal Server
HP NetServer
Microsoft Terminal Server Clients
HP Vectra
HP J31218A HP J3128A
Windows NT
(6 workstations) Windows 95 (6 workstations)
Figure 5