The Art and Science of Sizing Exchange 2003 (Part 3)

If you missed the other parts in this article series please read:

 

 

 

But first… Some Global Catalog sizing info

 

Before talking about tools, let me just say a couple of words about Global Catalogs (GC). Although this article is about Exchange sizing, GCs are so critical to an Exchange infra-structure that I think they also deserve some careful planning. If you disregard global catalog sizing you may end up with a perfectly sized Exchange server that simply won’t perform as expected.

 

A global catalog is required in each domain that contains Exchange servers, since this particular server is critical for some Exchange services (including log on, group membership, store services) and access to the global address list (GAL).

 

Consider the following when placing global catalog servers:

 

 

  • All Exchange servers and users should have fast access to a global catalog server. Verify that the DSAccess list only contains local DC/GC servers.
  • There should generally be a 4:1 ratio of Exchange processors to global catalog server processors, assuming the processors are similar models and speeds. However, depending on your situation, higher global catalog server usage, a large Active Directory, or large distribution lists can necessitate more global catalog servers.
  • Use the/3GB switch on global catalogs that have more than 1GB of RAM. It will increase the JET cache from 512MB to 1GB, so you’ll have more AD objects in memory.

 

And now, enter the tools

 

This article will not provide step-by-step instructions on how to use the mentioned tools, it just intends to reference them so that you know what’s available to help you.

 

HP Storage Planning Calculator

 

The HP Storage Planning Calculator for Microsoft Exchange Server 2003 is a fantastic tool that addresses several of the most critical areas of Exchange server design and provides guidelines for disk, storage controller and enclosure planning.

 

 


Figure 1: HP Storage Planning Calculator

 

This free tool provided by HP (a registration may be needed) eases the burden of planning the storage of an Exchange 2003 server. The nice piece of software takes into account factors like capacity, performance, backup time, restore time, user profile and it can even generate a pricing breakdown report!

 

The tool is designed to allow for a variety of what-if scenarios, comparing RAID levels, disk sizes, storage array controllers, enclosures, backup topology and tape drives.

 

Make sure you also visit the site HP ActiveAnswers for Microsoft Exchange Server, where you can find other tools and guidance documents that will help you with many Exchange related tasks.

 

System Center Capacity Planner (SCCP)

 

One of the most exciting solutions coming from Microsoft this year is System Center Capacity Planner 2006, a tool that can help to effectively size a Microsoft Exchange Server 2003 (and also MOM 2005). By using modeling technology, System Center Capacity Planner can help you to accurately determine the amount of hardware required to a successful deployment.

 

System Center Capacity Planner has the ability to prescribe a deployment architecture based on the characteristics of an organization such as the number of offices, the number of users and the expected usage patterns.

 

System Center Capacity Planner includes what-if analysis capability for hardware, software, and topology, so you will be able to experiment with alternative scenarios before going to production.

 

The next figures illustrate the model wizard and the produced results.

 

 


Figure 2: Capacity Planner wizard

 

 


Figure 3: Capacity Planner hardware results

 

User Monitor

 

Microsoft Exchange Server User Monitor will help you to better understand your user environment, by gathering real-time data. After collecting the data you can view several items, including IP addresses used by clients, versions and modes of Microsoft Office Outlook, and resources such as CPU usage, server-side processor latency, and total latency for network and processing with Outlook 2003 version MAPI.

 

There’s a great article on the MSExchange.org site with a detailed explanation of this tool: Using the Microsoft Exchange Server User Monitor (ExMon) tool.

 

Validation

 

No matter how you do the hardware calculations, the final step should always be the validation of the final system. In order to do that, Microsoft provides some appropriate tools that allow you to simulate some messaging loads and monitor the server behavior under stress:

 

 

 

 

Tool

 

Description

 

Available at:

 

Exchange Stress and Performance (ESP) 2003

 

A highly scalable stress and performance tool for Exchange 2003, to simulate large numbers of client sessions by concurrently accessing one or more protocol servers.

 

http://go.microsoft.com/fwlink/?LinkId=27881

 

Jetstress

 

Use Jetstress to verify the performance and stability of a disk subsystem. Jetstress helps verify disk performance by simulating Exchange disk Input/Output (I/O) load.

 

http://go.microsoft.com/fwlink/?LinkId=27883

 

Load Simulator (LoadSim) 2003

 

A benchmarking tool to simulate the performance load of MAPI clients. LoadSim allows you to test how a server running Exchange 2003 responds to e-mail loads..

 

http://go.microsoft.com/fwlink/?LinkId=27882

 

Network Monitor

 

Used mainly to diagnose issues related with server connectivity.

 

Part of the Operating System

 

Performance Monitor

 

A tool that helps establishing a baseline of performance and used for troubleshooting performance issues.

 

Part of the Operating System

Table 1: Exchange performance and troubleshooting tools

 

Exchange Stress and Performance (ESP) 2003

 

ESP 2003 simulates several arbitrary client sessions that are concurrently accessing one or more Exchange 2003 servers.

 

ESP provides modules that simulate client sessions over the following Internet protocols and APIs:

 

 

  • WebDAV (Microsoft Office Outlook Web Access)
  • Internet Message Access Protocol version 4rev1 (IMAP4)
  • Lightweight Directory Access Protocol (LDAP)
  • OLE DB
  • Network News Transfer Protocol (NNTP)
  • Post Office Protocol version 3 (POP3)
  • Simple Mail Transfer Protocol (SMTP)
  • Exchange ActiveSync
  • Outlook Mobile Access

 

Although ESP is similar to LoadSim, the first one is particularly adequate to validate deployments that use mobility features and internet protocols, while the second is indicated to simulate MAPI sessions.

 

Jetstress

 

Jetstress is a tool that helps verify the performance and stability of the disk subsystem prior to putting an Exchange server into a production environment.

 

Jetstress does this by simulating some heavy loads to the disk subsystem, just like if it was being used by a large number of users. You then use System Monitor, Event Viewer and the Eseutil tool together with Jetstress to verify that your disk subsystem meets or the established performance criteria.

 

Jetstress enables you to do two types of tests:

 

 

  • Disk Performance Test runs for two hours and enables you to verify the performance and sizing of your storage solution.
  • Disk Subsystem Stress Test runs for 24 hours and enables you to test the server storage reliability over a longer time.
  • The best way to verify the integrity and performance of your disk subsystem is to run both tests.

 

In the latest download package for Jetstress there are two separate applications:

 

 

  • JetstressUI.exe (the graphical user interface (GUI) version of Jetstress named Jetstress 2004)
  • Jetstress.exe (the original command-line version of Jetstress)

 

Both versions can be used to successfully test the performance of an Exchange disk subsystem, but using the command-line requires expertise in specifying the parameters and analyzing performance results. Jetstress 2004 and its graphical user interface reduces the complexity of configuring the test. Additionally, it facilitates the analysis of results by producing a performance analysis report.

 

Although it is recommended that you run the tests in a non-production environment, in the real world sometimes it’s not possible to have a lab that mirrors the actual systems. So, my advice here is that you use the production server just before going live, but remember to format the used storage volumes at the end of the tests.

 

Load Simulator (LoadSim) 2003

 

Load Simulator 2003 (LoadSim) should be used when simulating MAPI clients (Outlook). When using LoadSim you can select different user profiles and thus you can validate if your server can handle the estimated load.

 

LoadSim should be used together with other tools, since it does not simulate the following factors that can affect your server capacity planning:

 

 

  • Incoming unsolicited commercial e-mail (aka spam) from the Internet;
  • Incoming SMTP mail flow from the Internet or other sites within your organization;
  • Use of non-MAPI protocols for account access, such as POP3 and IMAP4;
  • Use of mobile devices;
  • Public folder usage.

 

Network Monitor

 

Network Monitor is a tool provided with Windows Server 2003 (although is not installed by default) and Microsoft System Management Server (SMS), that can be used to monitor the network and capture data “off the wire”, thus helping to detected potential problems.

 

With Network Monitor you can perform the following tasks:

 

 

  • Identify abnormal network traffic patterns. For instance, it allows detecting client-server communication problems, finding workstations that makes a disproportionate number of requests and identifying unauthorized users on your network;
  • Capture packets directly from the network;
  • Display, filter, save and print the captured frames.

 

To install Network Monitor, open the Add/Remove Programs applet from Control Panel, and click Add/Remove Windows Components. Select Management And Monitoring Tools, click Details, select the Network Monitor Tools check box, and click OK.

 

For more detailed instructions on how to use this tool read this article: Analyzing Traffic With Network Monitor.

 

 

Note:
Whenever possible you should use the version of Network Monitor included with SMS, since the Windows Server version has some limitations, such as it doesn’t capture network data for client computers or other servers. If you don’t have SMS (2.0 or 2003) there’s a great open-source tool that has all the required features: Ethereal.

 

Performance Monitor

 

Performance Monitor is a valuable tool that comes with the Windows operating system that supports detailed monitoring of the system resources. This tool provides performance indicators that are organized hierarchically by object, counter, and (optional) instance, as follows:

 

 

  • Performance Object – The part of the computer you can monitor. Some of the most commonly used objects are Processor, Memory, and PhysicalDisk. When you install Exchange 2003, new objects are added to the performance object list.
  • Counters – The parts of the object you can monitor. For example, you can monitor the available bytes or page faults per second of the memory object.
  • Instances (optional) – Multiple objects or counters to monitor on the computer. For example, when you look at counters under the Physical Disk object, you see as many instances as the number of disks on that computer. You can choose to monitor only a specific instance or all of them.

 

Although there are many, many counters you can check on an Exchange server, there are some main counters that can immediately indicate if there is some kind of bottleneck:

 

 

 

 

Object

 

Counter

 

Suggested values

 

Memory

 

Available Mbytes

 

>50 at all times

 

Pages/sec

 

<10 average, <100 on spikes

 

Processor

 

% Processor Time

 

<80% average at peak times

 

System

 

Context Switches / sec

 

<50,000 at the 95th percentile

 

Proc Queue Length

 

<twice the number of virtual processors

 

Network

 

Bytes sent / sec

 

These values should be monitored for trends, particularly on SMTP and front-end servers.

 

Bytes received / sec

 

Current Bandwidth

 

>7 Mbps sustained indicates the need for dual-100 Mbit NICs

 

Physical Disk <database>

 

Average disk sec/read

 

<20 ms 95% of the time, spikes < 60 ms

 

Average disk sec/write

 

Physical Disk <databases, transaction logs>

 

Average disk sec/read

 

<10 ms 95% of the time, spikes < 50 ms

Table 2: Performance objects and counters

 

 

Tip:
If you’re not sure what counters to watch for, you can use the Performance Monitor Wizard, which simplifies the process of gathering performance monitor logs. It configures the correct counters to collect, sample intervals and log file sizes.

 

Conclusion

 

After reading this 3-part article you should be able to effectively size an Exchange server, by collecting the correct metrics and doing the calculations, or “cheating” and using one the sizing tools described in this last part. Either way you are now in possession of the knowledge that can prevent you from having unpleasant surprises when deploying a new system or troubleshoot performance problems on existing servers.

 

If you missed the other parts in this article series please read:

 

 

 

Additional Reading

 

“Planning an Exchange Server 2003 Messaging System”
http://go.microsoft.com/fwlink/?LinkId=21766

 

“Optimizing Storage for Exchange Server 2003”
http://go.microsoft.com/fwlink/?linkid=49324

 

“Exchange Server 2003 Performance and Scalability Guide”
http://go.microsoft.com/fwlink/?LinkId=28660

 

“Troubleshooting Exchange Server 2003 Performance”
http://go.microsoft.com/fwlink/?LinkId=22811

 

Leave a Comment

Your email address will not be published.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Scroll to Top