Compaq BL10e - HP ProLiant - 512 MB RAM Technical manual

ISS Technology Update Volume 8, Number 2
1
Understanding and interpreting the VMmark benchmark..........................................................................1
Expected life of the HP Smart Array BBWC upgrade module battery pack ................................................4
Meet the Expert—Mitch Wright ..................................................................................................................5
Quick tip: Customizing the iLO Text Console ..............................................................................................6
Recently published industry standard server technology papers ...............................................................9
Contact us .................................................................................................................................................. 9
Understanding and interpreting the VMmark benchmark
Introduction
Using virtual machine technology to run multiple virtual servers on a single physical server continues to grow in importance as
the IT industry searches for ways to more efficiently use hardware compute resources in the data center.
Traditional benchmarks are typically designed to test the performance and scalability of a server performing a single large-scale
task such as database serving or web serving while running under a single operating system. Results of these types of
benchmarks, however, do not provide a reliable indication of the server‘s ability to scale when running multiple virtualized
systems in a virtual machine environment.
VMmark, which was developed by VMware and introduced in 2007, was the first and is still the dominant benchmark
designed specifically to measure and compare virtualized system performance for servers.
Structure of the VMmark benchmark
As a VMware product, the VMmark benchmark works solely with the VMware ESX hypervisor; therefore, it cannot be used to
measure the relative performance of other VMM solutions such as Windows Server 2008 Hyper-V or Citrix XenServer. VMmark
can be used to gauge the relative virtualization performance of different server platforms when running the ESX hypervisor.
The Tile –VMmark’s basic unit of work
VMmark uses a set of six virtual machines, each running a particular operating system and application workload, to define its
basic unit of work –the tile. Table 1 shows a detailed breakdown of the operating systems and common data center
applications that comprise each of the six virtual machines in a tile.
ISS Technology Update
Volume 8, Number 2
Keeping you informed of the latest ISS technology

ISS Technology Update Volume 8, Number 2
2
Table 1-1. VMmark workload summary per tile
Workload
Application
Virtual Machine Platform
Mail server
Exchange 2003
Windows 2003, 2 CPU, 1 GB RAM, 24 GB disk
Java server
SPECjbb2005-based
Windows 2003, 2 CPU, 1 GB RAM, 8 GB disk
Web server
SPECweb2005-based
SLES 10, 2 CPU, 512 MB RAM, 8 GB disk
Database server
MySQL
SLES 10, 2 CPU, 2 GB RAM, 10 GB disk
File servers
dbench
SLES 10, 1 CPU, 256 MB RAM, 8 GB disk
Standby server
None
Windows 2003, 1 CPU, 256 MB RAM, 4 GB disk
To ensure the benchmark‘s standardization, VMmark specifies how each virtual machine is configured, including memory, disk,
and number of CPUs. This is also shown in Table1.
Client systems are also required to drive the workloads on the tile of the server under test. Each tile requires a separate client
system, and the configuration and software to be run on the client is also strictly defined.
Calculating a Tile’s performance
A VMmark test run is designed to last for three hours. During this time, the system collects metrics on the performance of each of
the five virtual machines in the tile that are actually running workloads. The results for each virtual machine are then compared
against those obtained when running a single tile on VMmark‘s defined reference system to generate a normalized
performance ratio for each virtual machine. Finally, a geometric mean of these five numbers is taken to arrive at the
performance score for the tile. Because the performance numbers are taken as ratios against the VMmark reference platform,
tile performance scores tend to gravitate around one. In fact, since the reference system for all VMmark results is defined as an
HP ProLiant DL580 G2, the performance score for a DL580 G2 running a single tile is exactly 1.00.
Running multiple tiles
The VMmark benchmark can be set up to run a single tile or multiple tiles. Deciding how many tiles a given platform is capable
of running is an empirical determination; however, platforms with more processors and more cores can generally run more tiles.
When running more than one tile, the performance results from all of the tiles are totaled to generate an overall performance
score for the test. VMmark scores are always represented by two numbers –the aggregate performance scores of all the tiles
and the number of tiles run. A typical VMmark score for a system would be stated as 10.00 @ 7 tiles.
Interpreting VMmark results
Interpreting VMmark results requires looking at both numbers within the score. Clearly, when comparing two systems tested with
the same number of tiles, the system with the higher performance score (left number) has achieved a better result. Overall, the
ability to run more tiles is the more important metric since it indicates a platform‘s ability to run more concurrent workloads than
other systems. Also, higher performance scores will almost always correlate with higher numbers of tiles.
Because it is focused on virtualization performance, VMmark is essentially a processor and core bound benchmark. Systems
with greater numbers of processors and cores will generally outperform those with fewer. While the amount of physical memory
installed can make some difference, any installed memory greater than the five to six gigabytes required for each tile that the
processor complex is capable of supporting is of little or no benefit. Table 2, which contains the latest VMmark results for
several HP ProLiant servers, illustrates the difference that more processors and cores make in overall virtualization performance.

ISS Technology Update Volume 8, Number 2
3
Table 1-2. Recent HP ProLiant VMmark results
System
Processors
Cores
VMmark score
ProLiant DL385 G5p
2
8
11.28 @ 8 tiles
ProLiant DL585 G5
4
16
20.43 @ 14 tiles
ProLiant DL785 G5
8
32
27.71 @ 19 tiles
Certification of VMmark results
VMware maintains a review process for validating and approving VMmark benchmarks. These benchmark scores are
published on the VMware website: http://www.vmware.com/products/vmmark/results.html
The future for virtualization and virtualization benchmarks
Updates to the VMmark benchmark
The VMmark benchmark can be run on different versions of the ESX hypervisor, which will influence the results. When
evaluating VMmark results, it is always important to compare results that use the same software platform. As of February 2009,
the latest released platform from VMware is ESX v3.5.0 Update 3.
VMware has also released VMmark 1.1 in August 2008. VMmark 1.1 changes the operating systems for two of the virtual
machines in the VMmark tile. The Java server and Database server virtual machines now use the 64-bit versions of the
Windows and Linux operating systems respectively.
Work on an industry standard virtualization benchmark
The Standard Performance Evaluation Corporation (SPEC), which is responsible for developing and maintaining standardized
benchmarks used in the industry, is committed to developing standard methods of comparing virtualization performance. The
SPEC Virtualization Committee, which began as a working group in late 2006, is currently working to create an industry
standard benchmark that can be used to evaluate virtualization performance for data center servers. While a specific release
date for this benchmark has not yet been set, the SPEC Virtualization Committee‘s efforts are readily available at
http://www.spec.org/specvirtualization. Both HP and VMware, along with many other leading companies, are members of
SPEC as well as participants in the Virtualization Committee.
Additional resources
For additional information on the topics discussed in this article, visit the following websites:
Source
Hyperlink
VMware VMmark homepage
http://www.vmware.com/products/vmmark/
SPEC Virtualization Committee
homepage
http://www.spec.org/specvirtualization/
Overview of the VMmark
benchmark on HP ProLiant servers -
white paper
ftp://ftp.compaq.com/pub/products/servers/benchmarks/VMmark
_Overview.pdf
HP VMmark page
http://www.hp.com/go/vmware/vmmark

ISS Technology Update Volume 8, Number 2
4
Expected life of the HP Smart Array BBWC upgrade module battery pack
The Smart Array 128-MB Battery-Backed Write Cache (BBWC) upgrade module uses a nickel metal hydride (NiMH) battery
pack to retain write data if power to the system is interrupted. Two important battery pack characteristics are its charge
capacity and battery life. Charge capacity, expressed in milliampere-hours (mAh), determines how long—in days—the BBWC
module can retain data. A new, fully-charged battery pack typically provides 5 to 6 days of data cache retention. However, the
data retention time decreases with higher operating temperature and increasing battery age. The battery pack is designed to
provide 3 days of data cache retention at 80 percent of its rated capacity. Data retention time is also affected by the size of the
cache. Therefore, larger caches (256 MB and 512 MB) include up to three battery modules to retain data up to 3 days.
Note
The data retention time can be extended any time the server's auxiliary power is available during system power down.
As the battery pack ages, the battery capacity decreases in a roughly linear manner (Figure 1-1). Battery aging is accelerated
by high server internal temperatures. In a worst case scenario with the sever operating at its maximum temperature of 60°C
(140°F), the battery pack has a life expectancy of 3 years, after which it should to be replaced. If the server‘s inlet temperature
is maintained at 24°C (75°F) or less, then the server‘s internal temperature should remain below 60°C, and the battery pack
may last more than three years. However, HP recommends replacing the battery pack every 3 years.
Figure 1-1. The Smart Array BBWC upgrade module is designed to provide 3 days of data retention at about 80 percent of its rated
capacity. Normally, the battery pack can last for more than 3 years. However, high system temperature accelerates aging and decreases
battery life.
Additional resources
For additional information on the topics discussed in this article, visit the following websites:
Resource
URL
HP Smart Array E200 Controller User Guide
http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual
/c00688683/c00688683.pdf
HP Smart Array 6i Controller User Guide
http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual
/c00217854/c00217854.pdf

ISS Technology Update Volume 8, Number 2
5
Meet the Expert—Mitch Wright
Mitch Wright is a Technologist/Sr. Engineer for ASIC and Chipset Enablement in
the Technology Solutions Group-Industry Standard Servers (TSG-ISS). Mitch works
with product teams, system architects, and vendors to develop high-quality, highly
reliable ProLiant products on schedule. ―The key,‖ says Mitch, ―is to meet the
needs and requirements of our customers.‖
According to Ken Jansen, his manager, Mitch is fueled by his ―passionate
inquisitiveness. Mitch is always looking for new and better technologies to solve
our customers‘ problems, and he is good at discerning promising technologies
from those that aren‘t or those that will take longer to develop. Mitch also
evangelizes new technologies within the company to ensure that they reach the
customers.‖
Mitch‘s hands-on approach translates into his hobbies. He likes to design and
build wood furniture. He says, ―Working with your hands keeps you humble.‖
Mitch and Jenny (married for 21years) have three children: Grace (20), Claire
(14), and Alex (11).
Below are excerpts from an interview with Mitch.
Why did you decide to become an engineer?
From an early age, I had a huge curiosity about
how things work and I have been on a quest
ever since. Growing up, I always worked with
my dad in the garage and he continually
encouraged me to invent stuff and build it
myself. His only requirement was I use what was
readily available. So the build versus buy
question always ended the same way…build it.
What is your most interesting research or invention?
I have had many projects that are interesting for different reasons. For example, the ProLiant 8000 [first Compaq 8-processor
server] and the DL585 G1 [first ProLiant Opteron-based server] were both significant products. I have seven patents, but I am
proudest of my first patent and sole-inventor patent. Also, I worked as the Lead Technologist in the Physics Research
Instrumentation Lab at the Super Collider Lab from 1991-94. It was a once-in-a-lifetime experience because I worked with some
of the world‘s brightest physicists.
How much customer input goes into your product designs?
Customer requirements are my primary concern, whether designing computers for the data center or for scientific instruments. In
the last few years, I have been fortunate not only to work with vendors but also to interface with some large customer accounts.
I also interface with presales engineers at the HP Tech Tours hosted by the Technology Communications group. These
interactions provide a unique opportunity to understand customer challenges and requirements.
How do you drive customer input into the design of your products?
We work hard with our vendors to get the features right and to ensure that the components work as designed. In the ASIC and
Chipset Enablement group, we achieve this by driving industry standards and testing for compatibility and reliability with a
suite of tools developed over many years at HP.
Name: Mitch Wright
Title: Technologist/Sr. Engineer
Years at HP: 12
University: Texas Tech University
U.S. Patents: 7
Awards: TSG 2006 MVP
Technical Papers: In search of
Acceleration, High Speed Digital Design:
Moving from 33MHz to 133MHz PCI-X
An inventor at age 12

ISS Technology Update Volume 8, Number 2
6
Quick tip: Customizing the iLO Text Console
The iLO Text Console (TEXCONS) feature was introduced with iLO 2 version 1.70 and marks the return of a text remote console
session that can be accessed from Telnet or Secure Shell (SSH). iLO 2 also supports using Telnet to connect to the Text Remote
Console. However, the data stream is not encrypted when using a normal Telnet connection. When using SSH, the data stream,
including authentication credentials, is protected by the encryption method supported by the SSH client and iLO 2. HP
recommends using SSH to connect to the iLO Text Console. As part of the default security policy, using Telnet is disabled.
Administrators must enable Telnet to allow access to the CLI, and Text Remote Console.
For more information about the security of the communication methods used by iLO 2, see the Integrated Lights-Out Security
technology brief at http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00212796/c00212796.pdf
Customization
Implementing the TEXCONS feature allows for configurability. In general these options do not need to be changed but may be
useful in some situations.
When starting TEXCONS, options and arguments can follow the command line interface (CLI) command ‗textcons‘in order to
customize the operation of the display.
Controlling Rate of Sampling
The speed option can be used followed by an argument indicating the number of milliseconds between sampling periods where
the iLO firmware examines screen changes. This can alleviate unnecessary traffic on long / slow network links, or simply
reduce the bandwidth used and the time consumed by the iLO management processor CPU. Reasonable values are between 1
and 5000 (1ms to 5 seconds).
>textcons speed 500
Controlling Smoothing
The iLO processor attempts to transmit data only when it has changed on the screen and has then become stable. If a line of
text is changing more quickly than iLO can sample it, the line will not be transmitted until it has become stable. As an example,
during an ‗ls –R‘ of a large file-system, the physical monitor will display text more rapidly than it can be interpreted. In the
Software Text Remote Console case the same is true. The data will be displayed rapidly, and is essentially indecipherable. In
this case, however, the data is being transmitted by iLO across the network –consuming bandwidth. The default behavior is
the smoothing as defined above. The smoothing feature can be disabled with the delay option and an argument of 10.
>textcons speed 500 delay 10
Controlling Foreign Language Keyboard Support
While performing TEXCONS, the iLO 2 processor, can perform the necessary character mapping between the client, telnet,
and the server. The default mapping is the USB 101-keyboard translation (or no translation). The other keyboard support
translation modes are listed below.

ISS Technology Update Volume 8, Number 2
7
Table 4-1: Keyboard support translation modes
Keyboard
Reference Number
US keyboard
0
British Keyboard
1
Belgian Keyboard
2
Danish Keyboard
3
Finnish Keyboard
4
French Keyboard
5
French (Canadian) Keyboard
6
German Keyboard
7
Italian Keyboard
8
Latin America
9
Norwegian Keyboard
10
Portuguese Keyboard
11
Spanish Keyboard
12
Swedish Keyboard
13
Swiss (French) Keyboard
14
Swiss (German) Keyboard
15
The option to control the translation is the xlt option and the argument should be the reference number indicated in the above
table.
>textcons speed 50 xlt 1
This enables TEXCONS at a sampling rate of 50 ms and performs translation as if the input keyboard being used has a British
keyboard layout.
Configuring Remote Console Hot-Keys
In order to allow additional special key sequences that cannot be duplicated by the remote console client, either due to a telnet
/ SSH limitation or due to the client OS intercepting the keyboard sequence and processing the special key sequence itself, the
Remote Console Hot-Keys configured for Remote Console will operate in TEXCONS.
These could include items such as
Linux ‗magic‘ SysRq key support
X windows CTRL-ALT-BACKSPACE support
or simply short-cut for complex key sequences
Control Character Mapping
In general, under the ASCII character set, CONTROL (All ASCII characters below Code 32) characters are not printable
(cannot be displayed). In the IBM codepage 437, these characters are printable as various ‗special‘ characters. These
characters may be used to represent arrows, stars, circles, and so forth. In order to operate successfully on the majority of
clients (that do not display codepage 437 characters), some of these characters are mapped to equivalent ASCII
representations.

ISS Technology Update Volume 8, Number 2
8
Table 4-2: Control character mapping
Character Value
Description
Mapped Equivalent
0x07
Small dot
* - asterisk
0x0F
Sun
* - asterisk
0x10
Right pointer
>
0x11
Left pointer
<
0x18
Up arrow
^
0x19
Down arrow
v
0x1A
Left arrow
<
0x1B
Right arrow
>
0x1E
Up pointer
^
0x1F
Down pointer
v
0xFF
Shaded Block
blank space
All other characters are left un-altered.

ISS Technology Update Volume 8, Number 2
Legal Notices© Copyright 2009 Hewlett-Packard Development Company, L.P. The information contained herein is
subject to change without notice. The only warranties for HP products and services are set forth in the express
warranty statements accompanying such products and services. Nothing herein should be construed as constituting
an additional warranty. HP shall not be liable for technical or editorial errors or omissions contained herein.
TC090207NL
February 2009
Recently published industry standard server technology papers
Title
URL
Using HP ProLiant Network Teaming software with
Microsoft® Windows® Server 2008 Hyper-V
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c016632
64/c01663264.pdf
Implementing Microsoft Windows Server 2008 R2 Beta
on HP servers
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c016395
94/c01639594.pdfThe CONCENTRA File Upload/Download Applet uses
Microsoft VM. The Microsoft Virtual Machine (Microsoft VM) is a software
engine that runs platform-independent Java applets. To enable Microsoft VM
(if it exists on your PC), follow these steps.
1. From the menu bar, go to Tools / Internet Options / Advanced tab.
2. Under Microsoft VM, check all the 3 boxes.
3. Under Java (Sun), uncheck the box.
4. Click [OK] to save the changes.
If Microsoft VM is not listed on the Advanced tab, it indicates that your PC
does not have this version of Java.
Industry standard server technical papers can be found at www.hp.com/servers/technology.
Contact us
Send comments about this newsletter to TechC[email protected].