Editor's note: Ready to set speed records and scale the highest peaks in Windows? In this three-part series, Chris Amaris and Kenton Gardinier -- authors of "Windows 2000: Performance Tuning and Optimization" -- offer the dos and don'ts of fine-tuning software performance. In this article, Amaris and Gardinier discuss switching to native mode, network monitoring, and Exchange and Active Directory storage. In part one, they covered planning and testing, tune replication, and schema management. In part three, they tackle Group Policy and traffic analysis.
Don't rule out switching to native mode.
Many are still unsure of the reasons to switch to native mode and have misperceptions on how it will affect down-level clients. Making the switch requires you to upgrade all of your NT 4 BDCs prior to going to native mode. Otherwise the BDCs will no longer be a part of the AD domain. Contrary to popular belief, native mode affects how well down-level clients can interact with the AD domain but it won?t prevent clients from authenticating to the domain.
The advantage of switching to native mode from mixed mode is that you remove all the NT 4 scalability limitations and enhances your ability to control the domain through GPOs. From a non-performance point of view, native mode also and gives you universal groups and group nesting capabilities.
Do proactively monitor the network.
Want to know how much disk space you'll need in six months or want to know when you'll need to upgrade a system's processing capability? These are just a few of the questions that you'll be able to answer once you begin to continuously monitor and catalog system performance. Proactively monitoring gives you insight into numerous system performance aspects including usage scenarios, potential problems and bottlenecks, workload characterizations, and much more. Many tools exist to assist in monitoring performance. Many tools are immediately at your disposal including the System Monitor and Task Manager. Microsoft also develops separate monitoring applications such as Microsoft Operations Manager (MOM) and Application Center that help to automate monitoring tasks. Also, there are a number of third-party tools that also provide additional functionality such as reporting.
Do separate out Exchange and AD storage components into separate drives and channels, such as OS, logs and database.
Microsoft application transactional databases that are based on the Jet engine technology use a combination of database files and logs to deliver both performance and reliability. They use a lazy write approach to the database files, while spooling transactions to the logs files quickly and committing the transactions once the database has been written to. This allows the complex and slow process of writing to the database to happen over time for increased performance while transactions are protected by the logs files.
However, this combination can be ruined by incorrect placement of the database and log files. To get the performance advantages of this, you need to have the databases be written to one drive subsystem and the log files written to another.
To get the optimal performance, place the database files on a RAID 0+1 or a RAID5 partition, the operating system page file on a RAID1 partition and the log files on a RAID1 partition. These should all be on separate channels or even separate controllers.
One large real estate conglomerate with a centralized messaging environment was finding that they could only scale their Exchange messaging servers to a maximum of 500 users before the users would start to see greater than 1 second response times from the messaging client. They had an average 1 second response time service level agreement, so they would deploy an additional server every time the response time would reach that threshold.
After doing a performance analysis of the Exchange server configuration, it was found that page files, database and log files were located on a single RAID5 drive. This was causing a bottleneck in the disk subsystem. The subsystem configuration was changed to take advantage of the multiple disk channels that the platform supported by creating separate partitions for the page file, the databases and the logs. No additional hardware or software was required. This configuration change resulted in a 4-fold increase in performance, allowing the same servers to host over 2000 users with a better response time.
For more information:
Read part one of this series.
Read part three here
Is your Win2000 performing at its peak? Learn how to control your environment and substantially improve system performance from authors Chris Amaris and Kenton Gardinier in this SearchWindowsManageability online event, Maximizing Windows 2000 Performance, Thursday, May 23 at 2 pm EDT.
About the authors: Chris Amaris is CTO and Kenton Gardinier is senior consultant for Oakland, Calif.-based Convergent Computing, which provides IT consulting and technical services. They are co-authors of the boo, "Windows 2000: Performance Tuning and Optimization."