Table of contents

Introduction ......................................................................................................................................... 3
HP Systems Insight Manager ............................................................................................................. 3
Customizing Systems Insight Manager .................................................................................................... 3
Microsoft SQL Server ........................................................................................................................ 3
System resources .............................................................................................................................. 3
Job cleanup for a large number of managed systems ........................................................................... 4
Remote Microsoft SQL Server (recommended) ....................................................................................... 5
Systems Insight Manager Sizer tool .................................................................................................... 5
Customizing Systems Insight Manager to your environment ........................................................................ 6
Creating a hierarchy: One master CMS with several regions ................................................................. 6
Master CMS: ............................................................................................................................... 6
Regionalize CMS for: .................................................................................................................... 6
Discovery related ............................................................................................................................. 6
IP range ...................................................................................................................................... 6
Changing Timeouts and Retries .......................................................................................................... 8
System and event management .......................................................................................................... 8
Customizing the query system ......................................................................................................... 8
Customizing the task system ......................................................................................................... 11
Tuning Microsoft Windows Server 2008 .............................................................................................. 12
Using Performance Monitor to optimize Microsoft Windows 2008....................................................... 12
Configuring the logging feature ................................................................................................... 13
Examining the log data ............................................................................................................... 13
Performance Analysis of Logs (PAL) tool ............................................................................................. 13
Virtual memory paging ................................................................................................................... 14
Tuning disk input and output ............................................................................................................ 15
Disk tuning RAID 5 performance ................................................................................................ 15
Tuning a single disk .................................................................................................................... 15
Additional tips: .......................................................................................................................... 15
Configuring the network.................................................................................................................. 15
Microsoft Windows Server 2008 performance tuning ......................................................................... 16
CPU configuration ....................................................................................................................... 16
Systems Insight Manager memory requirements ............................................................................. 16
System Page file ......................................................................................................................... 17
Disk subsystem ........................................................................................................................... 17
Network subsystem ..................................................................................................................... 17
TcpTimedWaitDelay .................................................................................................................... 17
Tuning Microsoft SQL Server ............................................................................................................ 17
Optimizing Microsoft SQL Server performance with Microsoft SQL Server components ........................... 18
Worker threads .......................................................................................................................... 18
Configuring HP SIM for optimal performance
in large environments
2
Lazy Writer ................................................................................................................................ 19
Log manager .............................................................................................................................. 19
Moving tempdb location ................................................................................................................. 20
SQL Server minimum and maximum memory ..................................................................................... 20
PAE and AWE ............................................................................................................................... 20
PAE .............................................................................................................................................. 21
User-mode address space on 32-bit x86 ........................................................................................ 21
Configuring Microsoft Windows Server 2008 x32 to use PAE .......................................................... 22
Using AWE memory in Microsoft SQL Server 2008 x32 to access more than 2 GB memory ................. 22
Microsoft SQL Server Express ....................................................................................................... 23
Configuring Microsoft SQL Server 2008 x32 to use AWE memory .................................................... 23
Using AWE memory.................................................................................................................... 24
For more information .......................................................................................................................... 25
Call to action .................................................................................................................................... 25

Introduction
HP Systems Insight Manager
HP Systems Insight Manager is a web-based management application that integrates existing management
technologies with the latest advances in Internet technology. It provides a proactive, easy-to-use, automated, cost-
effective solution for managing distributed systems.
Systems Insight Manager enables management of standards-based, distributed computing environments. By
enabling browser access to its components, Systems Insight Manager enables you to manage systems and groups
of systems, anywhere, any time.
HP recognizes the challenges faces in managing distributed enterprise systems. The growing mix of networks,
computer platforms, virtualization, applications, and databases makes this task complex, especially in an era of
shrinking budgets. Ideally, you want to manage the availability of all these distributed resources with a minimal
impact to your network and system resources.
Systems Insight Manager collects two types of data:
Asynchronous events
Polled data
An asynchronous event refers to any communication which is initiated by a managed system to notify the Central
Management Server (CMS) of relevant changes in status.
Each polling type imposes consequences on the system and the network. In general, you want to enable
asynchronous events to carry the burden of problem detection because SNMP traps, WMI events, and DMI
indications are short messages that only affect network utilization periodically, and are initiated when systems
detect a problem.
Polling, on the other hand, causes regular, periodic traffic consisting of a request for status from the Systems
Insight Manager server, followed by responses from the polled systems. The more systems that are polled and the
more frequently they are polled, the more network resources are consumed to managed the environment. This
white paper describes how you can balance the consumption of these network resources with optimum health of
the systems in the network. For example, it is critical to monitor closely servers that perform critical functions, but
not each desktop or notebook.

Customizing Systems Insight Manager
Because each network is unique, you might need to customize Systems Insight Manager so that it provides the
most productivity while using the fewest network and system resources. You can customize Systems Insight
Manager to better fit your needs in several ways (setting timeouts, altering tasks and queries, and adjusting
Cluster Monitor, and so on).
Microsoft SQL Server
While Microsoft SQL Server (2008 or later) is self-tuning, you must perform several actions to improve
Systems Insight Manager performance. SQL Server can be adjusted and tuned in many of the same ways. Use the
information in this document to determine how to tune SQL Server for your system, network, and application
needs.
System resources

Systems Insight Manager system resources are divided into the following categories:

CPUThe type, number, and speed of processors needed to support a given workload
DisksThe number of drives needed to support the input/output requirements or the amount of storage space (in
MB) required to hold the data
MemoryThe amount of memory needed to support a given workload
NetworkThe number and type of network interface cards (NICs) needed to support workload demands
Ideally, you set the CPU as the limiting resource because processing power provides the most scalable solution.
However, this setup requires an established balance among the remaining resources (disk, memory, and
network). After this balance is achieved, adding more memory or additional processors does not necessarily
improve system performance. Increasing CPU speed is the best way to improve performance. However, the
limiting resource might be application-related, not hardware-related.
Job cleanup for a large number of managed systems
When running Systems Insight Manager in an environment that contains a large number of managed systems, the
completed job output might exceed several thousand files. This output can result in HP SIM consuming large
amounts of memory, task page GUI slowdown, or out of memory errors to requiring you to restart Systems Insight
Manager.

If you experience such issues, you can implement the following to alleviate memory and disk space consumption.
Users can adjust the retention values for completed tasks, which can be altered to reduce resource consumption.

The thresholds for task results data retention are located on the Task Results Settings page (Figure 1).
To view the Settings page, from the S menu, select Tasks and LogsView Task ResultsCustomize. You can fine-
tune historical task retention by adjusting the values to achieve the desired balance of interface performance with
the need to retain task results.


Figure 1: Task Results Page



When you need to manage up to 5,000 systems, HP recommends the following default settings. These settings
might not be ideal for every situation, but they serve as a starting point from which further adjustment might be
necessary depending on performance needs.
Table 1: Recommended thresholds for a large number of managed systems


Setting Recommended value

Scheduled Tasks:Number of Results 3

Scheduled Tasks:Number of Days 7

Run Now Tasks:Number of Results 100

Run Now Tasks:Number of Hours 8


Remote Microsoft SQL Server (recommended)

When the Systems Insight Manager database is collocated with Microsoft SQL Server on the same system,
performance degradation occurs because of input/output issues. It takes approximately 35% longer to process the
same number of systems when the database is located on a local server versus a properly configured Microsoft SQL
Server running remotely. For the remainder of this document, you can assume that Microsoft SQL Server and
Systems Insight Manager are located on separate remote servers, denoted as the management server, unless
otherwise specified.
Systems Insight Manager Sizer tool

The Systems Insight Manager Sizer tool assists systems administrators in building highly available, high
performance HP SIM deployments running on Microsoft Windows, Linux, or HP -UX operating systems. Based on the
results of HP quality assurance and performance testing, the tool recommends server configurations based on
projected management workloads, installed management applications, and number of console users. The tool also
provides performance and configuration data designed to help systems administrators optimize the performance of
existing Systems Insight Manager deployments. HP recommends that you consult the Sizer tool when considering
CMS hardware and database configurations.
The Systems Insight Manager Sizing Tool is located at http://www.hp.com/go/insight/sizer.
Customizing Systems Insight Manager to your environment
Creating a hierarchy: One master CMS with several regions
When creating a hierarchy in an environment where systems can be broken down into region consider creating
one master CMS and several regional CMS.
To create a hierarchy:
1. Create the master CMS to collect the entire event that is important to the business: critical, major, and all
ProLiant events.
2. Install any plug-ins on the regional CMS.
You must install each plug-in based on the use to the specific region. For example, if Region A has a virtual
environment and Region B does not, only Region A has HP Insight Virtual Machine Manager software for
ProLiant (VMM PL) plug-in installed.
The following examples are uses for a master CMS and regionalized CMS:
Master CMS:
o List of all server and important systems
o For inventory, contract, and warranty
o Long poll cycles (daily)
o Daily data collections (off-production hours, by region)
o No event handling except for it regional CMS

Regionalize CMS for:
o Better event handlingfewer events
o Use for active plug-inVMM, VPM, PMP, VCRM


Discovery related

IP range
By default, Systems Insight Manager pre-populates the discovery IP range with the network subnets used by the
management server. You must modify this to represent the correct range of systems that you need to manage in
your group. The Inclusion and Exclusion ranges enable you to fine-tune the systems that you want to discover using
IP range pinging.
After Systems insight Manager is installed and the Discovery IP range is established, the discovery process must run
to discover the systems on your network. For an empty database, the maximum time needed to complete an IP
range scan for initial system detection is
(Numbc o] IP Addcsscs)(Rctcs+1)1mcouts
16
in seconds (a fully populated
network can take a shorter time).
(Numbcr o IP AJJrcsscs) can be calculated from the range you indicate in the Inclusion Range, less the IP
addresses in the Exclusion Range (if any). By default, (Numbcr o IP AJJrcsscs) is the 254 IP addresses for each
subnet the management server is connect to, using a class C network mask scheme. The Automatic Discovery
window shows the progress of discovery. System status polling and data collection are invoked when new systems
are discovered. Polling and data collection, like discovery, have retries and timeouts associated with them after
they are invoked. HP recommends that you customize the default tasks in HP SIM to represent the network of
systems being managed. For servers that require constant vigilance (like Microsoft Exchange servers), create a
Hardware Status Polling task (based on a collection containing the Microsoft Exchange server systems) to run every
5 minutes. For file and print servers, create a separate Hardware Status Polling task to run once every hour. For
desktops and portables, either change the polling frequency to once a day or week, or turn off Hardware Status
Polling for these types of systems. To edit and schedule this task, go to Task & LogsView All Schedule Task.

Figure 2: Hardware Status Polling Task selected to be edited





Figure 3: Hardware Status Polling Schedule Task Page




Suggested polling intervals for the two default Hardware Status Polling task intervals should be configured as
follows:
Table 2: Suggested polling intervals for Hardware Status Polling
Number of systems managed Hardware Status Polling for servers Hardware Status Polling for nonservers
Less than 500 Use default of 5 minutes Use default of 30 minutes
501 to 2000 Change to 30 minutes Change to 2 hours
2001 to 5000 Change to 1 hour Change to 4 hours or greater

8
Note:
Do not let status polling tasks overlap. Time the Server Polling
taskcompletion time by making sure the interval is longer than the
completion time.
Changing Timeouts and Retries
HP SIM uses Ping, TCP, and HTTP, WBEM, and SSH protocols to determine if a system is on the
network. During the communication phase of discovery, you can customize several items to conform to
your network environment including Ping, SNMP Retries, and Timeouts.

HP SIM is tailored, out-of-the-box, for a LAN configuration. In many companies, the network has both
LAN and wide area network (WAN) components. To ensure good communication over a WAN, you
must consider network latency.

First, customize Network Timeouts. If systems are not being discovered through pinging, or information
is not being collected for a specific system over a WAN using SNMP, you must increase the Ping
Timeouts setting.

To customize HP SIM to your environment, begin by investigating how other applications are tailored to
communicate over your network. You can try increasing the Timeouts by a factor of 20 percent. For
example, by default SNMP Timeouts are set to 5 seconds. First increase the value to (5 * 1.2) 6
seconds. If this still does not enable you to discover a particular system, increase it to (6*1.2) 8
seconds.

Note:
Setting timeouts greater than 10 to 20 seconds can cause the various polling
tasks in HP SIM to take much longer to complete. In a mixed system latency
environment, it is best to adjust timeouts and retries for the slower-
responding systems using groups (such as LAN/WAN) and not through the
default settings on the Discovery page.
After a system is discovered, you can view a collection that contains that system and drill down to the
System Page window. From this window, you can further customize SNMP timeouts for that specific
system. After you change the value for that system, all tasks use the new timeout (or retry) when
communicating with that particular system. If your network has latency (WAN) problems, timeouts is the
best communication item to fine-tune. If you are experiencing packet loss with low latency networks,
timeout retries is the best item to customize (keep the current timeout values). HP recommends that you
set the minimum number of retries to one. Zero retries can result in packet loss and false notification of
unreachable systems.
HP SIM is designed so that if a request is sent out and the reply is received after the timeout period, the
reply is ignored. Therefore, you must tailor timeouts before retries. Increment retries only in single
values. For example, if the default ping retries value of 1 is not adequate for your system, the next
attempt must be 2.


System and event management

Customizing the query system
The collections feature of HP SIM enables you to generate information about a subset of systems or a
group. You can then perform tasks on all of the systems in that group with one collection. Several
default collections enable you to see collected information from the database.

Two default queries, All Systems and All Events, produce a listing of all systems that have been
discovered by Systems Insight Manager, and a list of all events that have been received. Separating
events by what is important to your business needs is essential to maintain the performance of Systems
Insight Manager. When managing a large number of systems (greater than 1,000), you can segment
the systems according to importance, location, type, operating system, and more. Leverage the
customize collections or Advanced Search features to create different collections to meet your needs.
For example, to fine tune the events to collect selected ProLiant events instead of all events:
1. Select Advanced Search and search for specific event types with a critical and major severity
type.

Figure 4: Collection for Critical and Major events for specific ProLiant events


When you view collections of up to 5,000 items, Systems Insight Manager can seem slow
depending on the system resources that are available on both the host and the client because of
the amount of data returned from the database. Segmenting collections by responsibility of area of
expertise (like e-mail servers, database servers, file servers, print servers, task servers, and line of
business servers) provides greater functionality for viewing specific information while improving
responsiveness. This also enables you to set up tasks specific to those particular systems. In a
normal network environment, you can potentially receive a large number of asynchronous events.
Creating a number of smaller collections allows the query system to function more responsively
and allows greater segmentation for executing a specific task when events occur.
Note:
The Discovered System event is generated so that actions can be taken
when new systems are added or removed from the network.
10
You can clear events so that undesired items can be easily filtered out. This filtering can be
performed automatically by creating a collection to filter all events older than 14 days, then
creating a task to delete events from that collection in a week. The following example filters all
Informational Events older than 14 days.

Figure 5: A collection to filter all informational events greater than 14 days old


2. After you have created and saved your collection, apply the collection to the delete event task by
selecting OptionsDelete Events. You saved the previous collection as Information
Events Older Than 14 days under Events by Time.

Figure 6: Select the collection saved and then click Apply.


3. To schedule the task, click Schedule.


Figure 7: Select Schedule


4. In the Task name field, name the task. For instance, the following task is named Delete
Informational Events Older than 14 days. You scheduled the task to run periodically
and refined the schedule to run every 14 days at 6:00 a.m. All these options are configurable to
your business needs.

Figure 8: Configure Delete task to run every 14 days


Customizing the task system
The task system of Systems Insight Manager enables you to execute a specific task on a group of
objects (whether they are systems, events, or a combination of both). When collections are segmented
in large network environments, you have greater control and an improved ability to customize
Systems Insight Manager according to your needs.
Initially, Systems insight Manager is configured to execute all the various polling tasks according to a
defined default schedule. These tasks are tied to queries in the System Default area. Most of these
collections are configured for the All Systems collection criteria. In a large network, this
12
arrangement might not be practical, nor would it be likely that you would want to poll all of your
desktops or all your servers on the same schedule.
By segmenting the collections by areas of responsibility or specific functions of systems (Exchange
servers, clusters, and so on), you can configure Systems Insight Manager to poll these systems
appropriately. For example, you can poll your most important Exchange or SQL servers once per
minute, and all other servers in your environment, once every 30 minutes. You might want to poll
desktops and notebooks only once per day or even once per week (using the DMI Status Polling task).
To segment your tasks, change the collection to which the default tasks are tied. You can edit these
tasks and select a smaller collection set (like Systems by Name or All Systems within X
Range). As a starting point, you can edit the collection criteria for the system default collections so
that you are polling a reduced set of systems instead of being tied to the All Systems collection
criteria. At this point, you must create new tasks (Hardware Status Polling, Data Collection, Software
Status Polling, and so on) and tie them to other collections that cover a different set of systems.
By creating several like tasks (Hardware Status Polling or Data Collection) against several different
collections, you can manage your network exactly the way you need to without inducing any
unnecessary traffic.
Note:
The task system is a serialized system for like tasks. This means that if you
create two data collection tasks, the first task executes and runs to completion.
The second task, if you scheduled it to execute in the same time period,
remains queued until the first task has completed. This means that two data
collection tasks do not run at the same time. If you have two dissimilar tasks,
such as Hardware Status Polling and Data Collection, then they execute
simultaneously.
Tuning Microsoft Windows Server 2008
Using Performance Monitor to optimize Microsoft Windows 2008
You can find the Microsoft Windows Performance Monitor that was used for most tuning tests under
StartProgramsAdministrative ToolsPerformance Monitor. Use caution with the
Performance Monitor because many counters and all Microsoft Windows 2008 servers resources
are interrelated.
HP recommends that you turn on Performance Monitor and log all objects and counters so that you are
not missing one you want later. When setting up the sampling times, remember that performance logs
grow large over time. To avoid being overwhelmed with data, look at a few key counters, then when
they show problems, dig a bit deeper.
At a minimum, monitor the following objects and counters:
Memory Object
o Memory: Available Mbytes
o Memory: Pages per second
Logical Disk
o Percent Disk Time
o Dedicated disk or disk drive that houses the pagefile.sys
Logical Disk Object
o Percent Disk Time
o Disk Queue Length
o Disk sec/transfer

Network Interface Object
o Output Queue Length
o Bytes Total per second
Processor Object (for CPU monitoring)
o Percent Total Processor Time
o Queue Length
In general, if an application is running at over 70% CPU utilization and a queue is forming, dig deeper
to determine which system resource is holding back the performance of your Microsoft Windows 2008
system.
Configuring the logging feature
Note:
Please note that the Windows Performance Monitor is a very powerful and
feature rich application. In addition to the Help feature, there are many
articles available online regarding relevant settings and counters details.
1. In the Administrative Tools program group, start the Performance Monitor.
2. Click the logging toolbar button (the drum).
3. Expand Data Collector SetsUser Defined in the left panel tree, right click on User
Defined, and select NewData Collector Set.
4. Enter the name of your new data collector.
5. Select Create from a template, and then click Next.
6. Select the System Performance template, and then click Next.
7. The next page will prompt for a destination folder for data. Consider potential data file size when
choosing the destination folder, as you may want to log performance data to a separate location.
8. Click Finish.
You will note that your counter will be added to the list of User Defined counters in the Data
Collector Sets tree, which resides in the left pane of the Performance Monitor.
9. Click your new counter to load it in the right pane.
10. If you want to change any aspect of the data collector, change the output file type, add/remove
counters, or change the frequency of logging, you can do so by right-clicking Performance
Counter and selecting Properties.
11. Once you are satisfied with your counter, select it in the tree, right-click, and select Start.
Examining the log data
Automated reports for performance counters can be generated and viewed by right-clicking the
desired counter and selecting Latest report. The Systems Performance Report displays an overview,
as well as detailed information regarding all selected counters. It can be a good starting point when
diagnosing performance bottlenecks on your system.
Performance Analysis of Logs (PAL) tool

When performing performance analysis and tuning activities, it might be helpful to employ the PAL tool.
PAL can analyze Windows Performance Monitor log files utilizing common threshold values, and
present an HTML report, which charts the performance counters and displays alerts when those
thresholds are exceeded. PAL is not a replacement for methods of performance analysis, but it can
provide meaningful information quickly and potentially save time when focusing on performance
14
optimization. You can find the tool at http://www.codeplex.com/PAL. It is open source and is released
under the Microsoft

Public License (MS-PL).
Virtual memory paging
The working set of a process is where its current code and data are physically located in RAM. When
sufficient RAM is available and a process requests information that is not currently in RAM, the virtual
memory manager in Microsoft Windows Server 2008 leaves the current working set in RAM and
retrieves the additional information from disk. This action is a soft page fault.
When RAM resources become scarce (either from the working sets of other processes or because of
the RAM that the file system cache is using), the virtual memory manager moves older pages from
RAM to the paging file system on the disk drive. This process is a hard page fault.
When a hard page fault occurs, the virtual memory manager has, in essence, stolen some RAM. In
other words, it has trimmed the working set of another process to fulfill a request from the process that
is currently running. Occasionally, paging that results from hard page faults is acceptable, but if this
type of paging occurs excessively, system resources become unbalanced and a memory bottleneck
forms.
To determine whether your system is paging, use the Performance Monitor to observe the relationship of
the metrics. If the pages-per-second counter is high (greater than 50) and growing compared with your
baseline, then memory is becoming bottlenecked. If the available bytes counter is also decreasing and
the disk drives that house the Pagefile.sys files are busy (marked by an increase in Percent Disk
Time, Disk bytes per second, and Average Disk Queue Length), you have a memory bottleneck.
You can improve the system memory and paging performance when loads are heavy by spreading the
paging file across two or more disks. This modification improves the overall paging file read/write
rates because more disks are available to process the paging file workload.
During installation, the operating system creates one Pagefile.sys file on the root (C:) drive.
To spread the load:
1. Review the disk metrics you gathered with Performance Monitor.
2. Select two disks that are under the lightest load.
3. Select Control PanelSystem.
4. Click the Advanced tab and the Performance button.
5. Click AdvancedVirtual Memory and create two new paging files, one on each disk.
6. After the new paging file systems are in place, remove the default Pagefile.sys on the root
disk.
As a guide to determining the Pagefile.sys sizes on the new disks, use Performance Monitor
to monitor the percent usage and percent peak usage counters of the paging file. Usually, you
create Pagefiles that are the same size on both disks, with initial sizes of a least the value shown
for percent usage and a maximum of at least the value for percent peak usage.
Sizing the Pagefile.sys correctly ensures that the server operating system does not waste
cycles creating larger Pagefile.sys files. If possible, dedicate two disk drives to the task of
containing the paging file systems. This approach guarantees that no other application or process
contend with the server operating system when the system needs the paging file system.
If the system begins to page at an unacceptable degree, use Performance Monitor and Task
Manager to isolate the applications or processes that are using excessive amounts of memory and
reduce the amount of memory allocated to them (if possible). If the application source code is
available, you can work with the application developers to improve overall memory performance.
When all tuning efforts fail to improve user response times related to a lack of memory, install
more memory into the system or distribute memory-intensive applications to the appropriate
number of additional management servers.

Tuning disk input and output
Disk tuning RAID 5 performance
Ensure that there are no other bottlenecks (CPU, NIC, or memory) before adding more disks. Typically,
the SCSI or Fibre Channel is not saturated, but rather there are not enough disks configured. Ensure
that you place the log files, indexes, and database files on different array sets.

The RAID stripe size is written to each system in the array or to the group of systems in the RAID set. In
other words, if you have a stripe size of 128 KB and four disks in a RAID 5 array, the stripe size is
divided across all the disks in the array. Each chunk is written to each drive in the array. So if you
have three systems in a RAID 5 array, three chunks are written. Logically, only two of the three drives
are receiving your data. The other drive is receiving the XOR parity data, which is rotated among the
disks.

For RAID 5, set the low-level raid format chunk size so that, when multiplied by the stripe width
(number of drives in the array minus one), the product is equal to the typical input/output size you use
from the server operating systems perspective.

For example, if you determine your input/output writes are occurring 8 KB at a time using Performance
Monitor and you have a three-disk RAID 5 set, chunk size must equal to 8 KB for two drives, or a 4 KB
chunk. If this is not possible, set the chunk size to the next larger size available.
Now, each 8 KB write fills an entire stripe set (user data) and the parity data is written on the same
stripe set on the last disk. Keep in mind that this parity info rmation rotates around the disks in the array.

Set the relationship between the RAID stripe size and the formatted sector/cluster size so that the
operating system input/output writes to the cluster size with the proper chunk size. Set the allocation
unit size of the NTFS partition to 8192.
Tuning a single disk
To optimize the performance of single, large hard disks in heavy throughput server environments, RAID
is the solution of choice, with throughput distributed over multiple drives and, more particularly, by the
application of large amounts (32 MB or greater) of write back cache on the hardware RAID controller.
However, many people are moving to more fault-tolerant configurations with full management server
mirroring and full fail-over for a more dynamic, high-availability solution. This solution deals with any
type of management server failure, not merely a hard disk failure. In this environment it is unnecessary
and costly to mirror complete RAID systems.
Additional tips:
Use only one disk partition per physical disk (lowers head movement)
Defragment the disk on a regular basis
Use the proper allocation size (match it to your workload)
Configuring the network
Generally, no specific tuning is required to get a NIC in Microsoft Windows Server 2008 to run at
full speed, but consider the following:
Do not use autodetect. Set your network system, clients, and management server NICs to the exact
settings you want. In many cases, autodetect provides less than optimal performance. Foregoing
autodetect also ensures that your network systems are working properly at the speed you desire. If
they are not, connectivity problems are likely to occur.
16
Ensure that nothing else is a bottleneck on the management server, client, or network. If your network
is not the limiting factor in your setup, then a fast NIC will probably not help. Ensure that the
application, CPUs, memory, and disks are not limiting your network performance.
Try a test case of your management server and clients on their own isolated LAN to see what kind of
throughput can be achieved.
Microsoft Windows Server 2008 performance tuning
Even though most performance tuning options are equally applicable to Microsoft Windows 2008,
there are additional performance aspects to consider when optimizing your Microsoft Windows server.
CPU configuration
Changing the CPU speed, the CPU cache size, or the number of CPUs in a web server can improve
performance significantly. Using a faster CPU or one with a larger cache always improve s the
performance of a web server with a CPU bottleneck. For example, on web servers that respond to a
significant number of dynamic requests and use encryption, increasing the number of CPUs, CPU
speed, or CPU cache size can be a very effective way to increase performance. Even though HP SIM
is typically not CPU bound, other system components, such as the operating system, Microsoft SQL
Server, or other applications, all benefit from the addition of another processor, improving overall
system performance. However, adding CPU resources to a C PU-limited Web server sometimes does
not improve performance. If you do not see much performance improvement for a highly dynamic site
when adding CPU resources, the problem might be with the design of the dynamic content or web
application. For static workloads, the CPU is unlikely to be the bottleneck; the usual culprit is the
network.

Memory
Web server performance is very sensitive to the amount of memory in a server. For example, Microsoft
Windows 2008 can cache high demand files in physical memory. By caching static files in memory,
the server can process requests more efficiently since disk input/output is eliminated (except for
logging). For the best performance, a web server should have enough memory to hold all static files. If
this is not possible, the disk subsystem becomes more critical.
Systems Insight Manager memory requirements
There are three main processes associated with Systems Insight Manager:
mxdomainmgr
mxdomainmgr is the main process that runs Systems Insight Manager. It is started by the primary
Systems Insight Manager service and is responsible for initializing and managing the core Systems
Insight Manager web application, partner code, and so on. This service runs the web server as well as
the bulk of the code that is Systems Insight Manager. It is not unusual to observe mxdomainmgr using
750 MB or more of memory.
Note that mxdomainmgr runs within a Java Virtual Machine process, which is restricted to the
maximum amount of memory available to any single process in Windows Server 2008 32-bit, namely,
approximately 1.6 GB. This restriction is because 32-bit Windows does not allow any single process to
allocate more than 2 GB of memory. The JVM heap must be allocated in a single contiguous block of
memory, and since Windows breaks up contiguous memory with additional libraries and so on, the
JVM never utilizes more than 2 GB of memory. Keep this in mind when considering system RAM
upgrades; though SQL Server benefits (as will other concurrent processes), the JVM does not benefit.
Using Server 2008 64-bit allows unrestricted maximum Java VM virtual address size.
mxdtf

msdtf is launched by the service. The main purpose of this process is to handle the audit logging and
SSH command line execution for MSA, SSA, and command launch tools. Memory for this process is
not nearly as high, usually less than 100 MB.
mxinventory
mxinventory is started as needed for WBEM and DMI data collection. When idle for longer periods it
shuts itself down for a short time. Note that for there might be two instances of this process started, one
for WBEM the other for DMI.
System Page file
If your configuration has two or more hard disk drives, HP recommends that you move the system page
file to a different drive than the one where the Microsoft Windows 2008 operating system is installed.
Splitting the page file across multiple disks also improve s paging performance.
Disk subsystem
The disk subsystem has very little effect on web server performance for static workloads that can be
cached in memory. For web servers that have less memory or that use disk input/output to generate
dynamic content, a RAID subsystem with at least four disks can improve performance substantially over
a configuration that uses single disks connected to a SCSI controller.
Network subsystem
The total bandwidth available to the web server spread across all of the network adapters in the server
sets the limit for the number of bits that a server can send or receive. For web servers on the Internet,
this is probably the most common bottleneck. Correspondingly, the network bandwidth limits the
request rate a server can handle. To determine the performance capabilities of a web server, you must
make sure that there is enough network bandwidth so that the tests reach the servers peak
performance. For instance, if a powerful server only has 100 Mb of bandwidth, this could be the factor
that prevents the server from performing better. If the CPU is the bottleneck, and there is still some
network bandwidth available, using a network adapter that supports the offloading capabilities in
Microsoft Windows 2008 can free CPU cycles to process more requests.
For more information regarding the tuning of Windows Server 2008, please consult the following
Microsoft document:
Performance Tuning Guidelines for Windows Server 2008
TcpTimedWaitDelay
When managing a large number of devices with Systems Insight Manager, the CMS opens many
network connections during the various polling and communications tasks. When a TCP connection is
initiated, used, and closed, a standard interval of time must pass between the time the connection is
closed and the resources are released. Once the connection is released, it can be used to create a new
connection.
When polling and communicating with a very large number of systems, there is a possibility that
opening new connections will be attempted even though the previous closed connections arent yet
available. In order to combat this, HP recommends that you change the TcpTimedWait delay from 240
seconds (0xfe) to 120 seconds (0x78). This is accomplished by locating the following registry key:
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\services\Tcpip\Parameters and changing the
TcpTimedWaitDelay dword value to 0x78.
Tuning Microsoft SQL Server
The performance of Microsoft SQL Server depends on the interaction of many factors, ranging from
the hardware that the system is running to the application coding techniques used. Microsoft SQL
Server provides methods and tools to tune Microsoft SQL Server for optimum performance. Keep these
18
principles in mind when you are tuning Microsoft SQL Server:
Plan for disk input/output subsystem performance.
The placement of the paging file, transaction logs, and system databases can have a big impact on
system performance and recoverability. Place log files and database files on different physical drives.
Microsoft SQL Server performance can be enhanced by locating log files and database files on
separate physical drives. The physical disk subsystem must provide a database server with sufficient
input/output processing power for the database server to run without disk queuing, which indicates
poor performance.
RAM is a limited resource
An integral feature of the database server environment is the management of RAM buffer cache.
Access to data in RAM cache is much faster than access to the same information from disk, but
RAM is a limited resource. If database input/output can be reduced to the minimum required set of
data and index pages, these pages stay in RAM longer. Too much unnecessary data and index
information flowing into buffer cache quickly pushes out valuable pages. The focus of performance
tuning is to reduce input/output so that buffer cache is best utilized. Memory is a cheap resource
always in demand. No amount of tuning can compensate for a lack of RAM.
Let Microsoft SQL Server do most of the tuning.
Microsoft SQL Server has been enhanced to create an auto-configuring and self-tuning database
server. Take advantage of the auto-tuning settings available. These settings help Microsoft SQL Server
run at peak performance even as user load and queries change over time.
Use the Create Index Wizard feature to create indexes for queries you add.
A key factor in maintaining minimum input/output for all database queries is to ensure that good
indexes are created and maintained. This wizard guides you through a series of steps to create
indexes, prompting you for input along the way.
Take advantage of Microsoft SQL Server Profiler and Index Tuning Wizard.
You can use Microsoft SQL Server Profiler to monitor and log a Microsoft SQL Server workload,
which can then be submitted to the Index Tuning Wizard to tune indexes for better performance.
Regular use of Microsoft SQL Server Profiler and the Index Tuning Wizard helps you optimize the
indexes, allowing Microsoft SQL Server to perform well with changing query workloads.
Monitor the management server with Windows Performance Monitor.
Regularly monitoring the management server, especially its disk utilization, reveals important clues for
troubleshooting performance problems. Use Microsoft SQL Performance Monitor before you have a
problem to establish a performance baseline for your management server. Microsoft SQL Server
provides counters that the Microsoft Performance Monitor can track. Counters that can help you
monitor disk activity include PhysicalDisk (Percentage Disk Time), PhysicalDisk (Average Disk
Queue Length), Microsoft SQL Server (Buffer Manager Page Reads/second), and Microsoft SQL
Server (Buffer Manager Page Writes per second). These disk input/output counters are available in
the Windows Server 2008 performance monitor tool.
Optimizing Microsoft SQL Server performance with Microsoft SQL
Server components
Worker threads
Microsoft SQL Server maintains a pool of Microsoft Windows operating system threads to service
batches of Microsoft SQL Server commands submitted to the database server. The total of these threads
(called worker threads) available to service all incoming command batches is dictated by the setting for
the sp configure option max worker threads. If the number of connections actively submitting
batches is greater than the number specified for maximum worker threads, the worker threads are
shared among connections actively submitting batches. The default setting (255) works well for many
installations.

Worker threads write out most of the dirty 8-KB pages from the Microsoft SQL Server buffer cache.
Input/output operations are scheduled by worker threads asynchronously for maximum performance.
Lazy Writer
Microsoft SQL Server Lazy Writer helps produce free buffers, which are 8-KB data cache pages that
contain no data. As Lazy Writer flushes each 8-KB cache buffer out to disk, it initializes the cache page
identity so that other data can be written into the free buffer. Lazy Writer produces free buffers during
periods of low disk input/output, so that disk input/output resources are readily available for use and
so that there is a minimal impact on other Microsoft SQL Server operations.
Microsoft SQL Server automatically configures and manages the level of free buffers. Monitor the
Microsoft SQL Server: Buffer Manager, Free Buffers object to ensure that the free buffer level remains
steady. Lazy Writer maintains the level of free buffers to keep up with the user demand for free buffers.
The Microsoft SQL Server: Buffer Manager, Free Buffers object should not drop to zero as this indicates
that there were times the user load demanded a higher level of free buffers than the Microsoft SQL
Server Lazy Writer provided.
If the Lazy Writer cannot keep the free buffer steady, or at least above zero, it might mean the disk
subsystem cannot provide Lazy Writer with the disk input/output performance that it needs to maintain
the free buffer level. You can compare drops in free buffer level to any disk queuing to confirm that
there is a disk subsystem problem.
One solution to the disk queuing problem is to add more physical disk drives to the database server
disk subsystem to provide more disk input/output processing power. The Microsoft SQL Server: Buffer
Manager, Lazy Writes/sec object indicates the number of 8-KB pages written to disk by Lazy Writer.
Monitor the current level of disk queuing in Performance Monitor by looking at the counters for (logical
or physical) Disk: Average Disk Queue or Current Disk Queue. The disk queue needs to be at a level
less than 2 for each physical drive associated with any Microsoft SQL Server activity. For database
servers that employ hardware RAID controllers and disk arrays, divide the number reported by disk
counters (logical or physical) by the number of actual hard disk drives associated with that logical drive
letter or physical hard disk drive number reported by the Disk Administrator program. Microsoft
Windows and Microsoft SQL Server are unaware of the actual number of physical hard disk drives
attached to a RAID controller. You should know the number of drives associated with RAID array
controller to interpret the Performance Monitor reports on disk queue numbers.
Monitor disk queuing on hard disk drives associated with Microsoft SQL Server data files to see if
Microsoft SQL Server is sending down more disk input/output requests than the disks can handle. If this
is the case, then more disk input/output capacity must be added to the disk subsystem to handle the
load.
Log manager
Like other major Relational Database Management System (RDBMS) products, Microsoft SQL Server
ensures that all write activity (inserts, updates, and deletes) performed on the database is not lost if
something interrupts the Microsoft SQL Server online status (power failure, disk drive failure, fire in the
data center, and so on). The Microsoft SQL Server logging process helps guarantee recoverability.
Before any implicit (single Transact-SQL query) or explicit (transaction that issues Begin Transaction,
Commit, or Rollback statements) transactions can be completed, the Microsoft SQL Server log manager
must receive a signal from the disk subsystem telling it that all associated data changes have been
written successfully to the associated log file. This rule guarantees the transaction log can be read and
reapplied in Microsoft SQL Server when the management server is restarted after an abrupt shut down
during which the transactions written into the data cache are not yet flushed to the data files. Flushing
data buffers are Checkpoint or Lazy Writer responsibility. Reading the transaction log and applying the
transactions to Microsoft SQL Server after management server stoppage is referred to as recovery.
Disks containing Microsoft SQL Server log files must have sufficient disk input/output handling capacity
for the anticipated transaction load because Microsoft SQL Server must wait for the disk subsystem to
complete input/output to Microsoft SQL Server log files as each transaction is completed.
20
For highest performance, you can use a caching controller for Microsoft SQL Server log files if the
controller guarantees that data entrusted to it is written to disk ev entually, even if the power fails.
Moving tempdb location
If you can add more physical drives to the system, HP recommends that you move tempdb locations to
different physical drives to avoid read or write delays associated with operating system and SQL Server
background activity.
By default, tempdb is placed in the same physical directory as the SQL Server binaries. To move
tempdb to another physical location, enter the following commands in SQL Server Management Studio
or an SQL Server prompt (note that <new location> represents the desired location of tempdb on the
new physical drive:

use master
go
Alter database tempdb modify file (name = tempdev, filename = <new

location>\tempdb.mdf)

go
alter database tempdb modify file (name = templog, filename = <new

location>\templog.ldf)
go


SQL Server minimum and maximum memory
When tuning SQL Server, it might be important to adjust the minimum and maximum amount of
memory that can be consumed. If Systems Insight Manager and SQL Server are running on the same
machine, it becomes even more critical as the operating system, Systems Insight Manager, SQL Server,
and other applications compete for RAM.
It might be difficult to know how much memory to allocate to SQL Server, especially as system
configuration becomes more complicated as noted above. On a server with 4GB of RAM that is also
running Systems Insight Manager, HP recommends that 50% of memory be allocated to SQL Server
maximum value. As the system runs, gather performance statistics by perfmon over an extended period
of time. These statistics can be analyzed to determine if it is necessary to adjust the minimum and
maximum values further.
The settings can be configured as follows (note that values are in MB):

sp configure 'min server memory', 1024
RECONFIGURE GO

sp configure 'max server memory', 2048
RECONFIGURE

GO

PAE and AWE
Physical Address Extension (PAE) refers to the feature of the X86 processors that allows them to address
more than 4 GB of memory. Previously, the X86 32-bit processors were only able to address 2
32
(4 GB)
of memory, however the addition of more address lines increased that amount to 2
36
(64 GB). All
additional address space is still unusable without operating system support, and since the 32-bit virtual
addressing space is unchanged, it must use page tables to map the 4GB virtual space to the 64GB
physical space. Applications are still limited to the traditional 4GB virtual space as well.
The method for monitoring disk queuing is different for Microsoft SQL Server log files than it is for
Microsoft SQL Server database files. You can use the Performance Monitor counters Microsoft SQL
Server: Databases database instance: Log Flush Waits Times and Microsoft SQL Server: Databases
database instance: Log Flush Waits/sec to view log writer requests waiting on the disk subsystem for
completion.

Advanced Windowing Extensions (AWE) is a method that the Windows operating system can make
more than 4 GB available to applications through system calls. To take advantage of AWE memory,
you must configure both Microsoft Windows 2008 x32 and Microsoft SQL Server.
PAE
PAE is the added ability of the IA32 processor to address more than 4 GB of physical memory. Note
that this section does not apply to 64-bit versions of Windows.
Note:
This applies specifically to Windows Server 2008 Enterprise Edition.
Windows Server 2008 x32 Standard Edition cannot address above 4GB
of RAM.
The following table lists the maximum physical memory that a Windows version can recognize, with the
PAE switch enabled (note that Windows Server 2008 R2 does not offer a 32-bit version):
Table 3: Maximum supported memory for Server 2008 R2 versions

Version Limit on X64
Windows Server 2008 R2 Datacenter 2 TB
Windows Server 2008 R2 Enterprise 2 TB
Windows Server 2008 R2 Foundation 8 GB
Windows Server 2008 R2 Standard 32 GB
Windows HPC Server 2008 R2 128 GB
Windows Web Server 2008 R2 32 GB


Table 4: Maximum supported memory for Server 2008 versions
Version Limit on X86 Limit on X64
Windows Server 2008 Datacenter 64 GB 1 TB
Windows Server 2008 Enterprise 64 GB 1 TB
Windows Server 2008 HPC Edition 128 GB
Windows Server 2008 Standard 4 GB 32 GB
Windows Small Business Server 2008 4 GB 32 GB
Windows Web Server 2008 4 GB 32 GB

Typically, a process running under Windows Server 2008 x32 can access up to 2 GB of memory
address space with some of the memory being physical memory and some being virtual memory. The
more programs (and, therefore, more processes) that run, the more memory you commit up to the full 2
GB of address space.
When this situation occurs, the paging process increases dramatically and performance might be
negatively impacted. The Windows Server 2008 x32 memory managers use PAE to provide more
physical memory to a program. This reduces the need to swap the memory of the page file and results
in increased performance. The program itself is not aware of the actual memory size. All the memory
management and allocation of the PAE memory is handled by the memory manager independently of
the programs that run.
User-mode address space on 32-bit x86

22
Previously with Windows Server 2003 x32, system memory was divided; 2GB was allocated to user-
mode processes and 2GB to kernel-mode process address space. If allocating more memory to user-
mode space was desired, the /3gb switch could be placed in the boot.ini file. With Server 2008, the
IncreaseUserVA boot option can be used to increase user-mode address space, thus making more
memory available to user processes. This setting is applied using the bcdedit tool as follows:
>Bcdedit /set IncreaseUserVA 3072

Configuring Microsoft Windows Server 2008 x32 to use PAE
Microsoft Windows Server 2008 does not support setting boot options in the boot.ini file. Instead, you
must employ the BCDEdit utility to configure PAE options. (The PAE option applies only to 32-bit
Windows versions.) If the system is configured for hot-add memory devices with address ranges
beyond 4 GB, Windows Server 2008 automatically configures itself with the PAE option enabled.
To enable PAE, enter the following command:

>bcdedit /set {current} pae default (or ForceEnable)


Before configuring PAE in Microsoft Windows Server 2008, consult the BCDEdit reference page at:
http://msdn.microsoft.com/en-us/library/aa906211.aspx.
Using AWE memory in Microsoft SQL Server 2008 x32 to access more than 2 GB memory
AWE is not needed and cannot be configured on 64-bit operating systems. Though the Use AWE
option is available in Microsoft SQL Server 2008 64-bit edition, it is ignored. However, it is still
recommended that this option be selected.
Because many Microsoft SQL Server databases are huge, greatly exceeding 2 GB, more than 2 GB of
RAM is often needed to attain a 90% or greater buffer cache ratio. Because of this problem, Microsoft
Windows Server 2008 and Microsoft SQL Server 2008 support a feature called Address Windowing
Extensions (AWE). AWE is a way to allow Microsoft SQL Server 2008 x32 to access more than 2 GB
of memory.
In most Microsoft SQL Servers, the biggest performance bottleneck is input/output. First, the purpose of
a database is to store data, and data is stored on disk arrays, requiring input/output to retrieve the
data from the disk array before it can be put into RAM for use. Second, disk access is the slowest
component of any Microsoft SQL Server. Combined, these two factors contribute to this common
bottleneck.
To help reduce input/output bottlenecks, Microsoft SQL Server includes a buffer cache, located in
RAM, used to store the most recently accessed Microsoft SQL Server data. By caching data, Microsoft
SQL Server does not need to access disk input/output as often, helping to boost overall performance of
Microsoft SQL Server.
The greater the size of the buffer cache, the more data Microsoft SQL Server can store in RAM, and the
less input/output access is required. In an ideal world, all of the data in a database would be stored in
the buffer cache, greatly reducing disk input/output, and boosting Microsoft SQL Server's performance.
In some cases, this is possible, but in most cases, it is not.
For ideal performance, the buffer cache hit ratio (the amount of data in RAM that is being reused)
should exceed 90%. If it does, excessive disk input/output is generally avoided, helping performance.
But if the buffer cache hit ratio is less than 90%, then disk input/output might become a serious
bottleneck.
The easiest way to boost the buffer cache hit ratio in a Microsoft SQL Server is to allocate additional
physical RAM to the Microsoft SQL Server. Assuming you have not changed the default memory
configuration settings in Microsoft SQL Server it then automatically uses as much RAM as you add to
your server.


Microsoft Windows 2008 x32 is a 32-bit operating system. By default, it can support only up to 4 GB
of memory (2 GB for the operating system and 2 GB for user applications). In a practical sense, this
generally means that Microsoft SQL Server is limited to 2 GB of RAM. In many cases, this is more than
enough RAM for Microsoft SQL Server. See Configuring Microsoft Windows Server 2008 to use PAE.
Microsoft SQL Server Express
When installing HP SIM, you can install and use Microsoft SQL Server Express. There are important
considerations to keep in mind when choosing this option, especially when planning performance
tuning. SQL Server Express is limited to 1 GB of RAM, one processor, a 4 GB database size, and runs
in Windows on Windows (WOW) mode on a 64-bit platform. Therefore, there might not be a
significant benefit in upgrading system memory, CPU, or storage size when using SQL Server Express.
In addition, if Systems Insight Manager is managing many systems, events, or performing historical data
collection, HP recommends that you do not use SQL Server Express, and instead use a more capable
version of SQL Server.
Configuring Microsoft SQL Server 2008 x32 to use AWE memory
Even though Microsoft Windows 2008 x32 can use the AWE memory, you must configure Microsoft
SQL Server 2008 x32 appropriately.
AWE Support must be enabled: To enable AWE memory on Microsoft SQL Server, enter the following
commands:
sp configure 'show advanced options', 1
RECONFIGURE GO

sp configure 'awe enabled', 1
RECONFIGURE GO

After SQL Server restarts, the following message appears in the SQL Server error log: Address
Windowing Extensions enabled.
Because the awe enabled option is an advanced option, you must first permit advanced options to be
set by running the show advanced options option, as shown above.
IMPORTANT:
To use AWE memory, you must run the Microsoft SQL Server 2008
mssqlserver service under a Microsoft Windows 2008 account that has
been assigned the Microsoft Windows 2008 lock pages in memory
permission. Microsoft SQL Server Setup automatically grants the
mssqlserver service account permission to use the Lock Page in Memory
option. If you start an instance of Microsoft SQL Server 2008 from the
command prompt using sqlservr.exe, you must manually assign this right to
the interactive user's account using the Microsoft Windows Group Policy
utility, or Microsoft SQL Server will be unable to use AWE memory when
run from the command prompt.
Note:
If you are clustering Microsoft SQL Servers that use AWE memory, it is
important that all the servers in the cluster have the same amount of RAM
and have their settings configured identically, otherwise, a failover
operation may not be successful.
If you have successfully enabled AWE memory, the message Address Windowing
Extension enabled appears in the Microsoft SQL Server error log when Microsoft SQL
Server is started.

24
Using AWE memory
After both Microsoft Windows 2008 and SQL Server 2008 are properly configured, you can take
advantage of the benefits of your added RAM. If you have just added RAM to a new Microsoft SQL
Server, you must watch the buffer cache hit ratio to see the effect of the addition of more RAM. If you
still do not get a 90% or more buffer cache hit ratio after adding the extra RAM, you might need to
consider adding even more RAM, or redesigning your database so that it is more efficient.
Carefully monitor the paging on the server after first configuring AWE memory. Microsoft SQL Server
does not dynamically manage AWE memory. If you see that paging starts to become a problem, you
might want to reduce the amount of RAM Microsoft SQL Server can access by reducing the Microsoft
SQL Server max server memory setting. It might take some experimentation before you find the ideal
setting for the maximum memory Microsoft SQL Server should have. Ideally, paging should be virtually
nonexistent on a well-tuned Microsoft SQL Server.


For more information
www.hp.com/go/somewhere
HP Partitioning Continuum, HP, 2008
HP ENSAextended technical overview, HP, 2008
HP Utility Data Center Overview, HP, 2008

Call to action
www.hp.com/go/somewhere
HP Partitioning Continuum, HP, 2008
HP ENSAextended technical overview, HP, 2008
HP Utility Data Center Overview, HP, 2008




Technology for better business outcomes
Copyright 2010, 2011 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.
533904-003, April 2011