382 CHAPTER 9 IMPROVING PERFORMANCE Table 9.1: Basic

384 CHAPTER 9 IMPROVING PERFORMANCE Monitoring Exchange 2003 Once you ve established that your Windows server is running, you can move on to monitoring your Exchange 2003 components. Now, how do you know if your Exchange server is healthy? Is it OK so long as your users don t complain about slow access times to their mailboxes? Or maybe your Exchange server is doing fine as long as you are able to send and receive Internet e-mail messages. Although these may be factors in helping you determine whether the Exchange server is healthy, you shouldn t use these as guides to determine the performance and health of your Exchange 2003 server. Instead, use the Performance console and System Monitor, which allow you to track and view trends of your Exchange 2003 servers and the usage of the specific services and protocols that are enabled for the Exchange 2003 environment. Have you found a bottleneck on the system and would you like to ascertain which Exchange 2003 component is causing the most system resource usage? Again, it s time to use System Monitor. Exchange 2003 Performance Console Objects When Exchange 2003 is installed on a Windows server, more than 30 additional performance monitor objects and almost 1,000 additional counters are installed. You may use some of these objects only in rare circumstances (or never), while others may prove useful to you on a regular basis. From within the System Monitor, if you click the Performance Object drop-down list, you will get a listing of the performance objects available. Figure 9.8 shows what this list might look like. Figure 9.8 Selecting the System Monitor object to monitor Some of the objects included in Windows may also be useful when monitoring Exchangespecific technologies such as the SMTP counters. You can find a complete listing of the performance objects related to Exchange at www.somorita.com/ e2k324seven/perfobjects.doc. Not all of these performance objects will be found on all Exchange servers. Here are more detailed descriptions of the more common objects: Database Counters relate to overall database performance such as cache hits, log writes, and pages converted per second. There is a database instance for the Information Store and one for the search service.

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Low Cost Web Hosting services

382 CHAPTER 9 IMPROVING PERFORMANCE Table 9.1: Basic

MONITORING WINDOWS 2003 PERFORMANCE 383 TCP Segments Retransmitted/sec The number of TCP segments retransmitted per second as a result of network problems. In a perfect world, should be zero, but in reality, it should be less than 5% of the TCP Segments Sent/sec counter. You have a network problem. Your network is too busy or there is an unreliable link. Look at your network infrastructure to see if you can improve this. Network Segment % Network Utilization The percentage of network bandwidth used on this segment. The ideal value for this counter will vary from network type to network type, but it should be less than 30 40% on an Ethernet network. I have seen healthy switched networks run much higher. Break your network into small pieces or implement switching technology. Network Interface Output Queue Length The length of the queue for packets waiting to be transmitted. Values higher than 2 indicate a possible network bandwidth problem. Break the network into smaller pieces or implement switching. Also consider getting a faster network adapter, looking at the network adapter settings, or updating the device driver. Process % Process Time (lsass instance) Percentage of processor time that the local security authority is using. If this is a domain controller, this is the Active Directory process. On a domain controller, you can always expect to see this process using some CPU time. If it is sharing a server with Exchange, it should not use more than about 15%. If it is exceeding the amount you expect, add a processor or move some tasks to another server. System Context Switches/sec The combined rate at which all processors on the computer are switched from one thread to another. Context switches over 4000/sec might be cause for concern. If it is exceeding the amount you expect, add a processor, or move some tasks to another server. Table 9.1: Basic Counters for Monitoring Windows Performance (continued) Object Counter Explanation Desired Values How to Improve

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Low Cost Web Hosting services

382 CHAPTER 9 IMPROVING PERFORMANCE Table 9.1: Basic

382 CHAPTER 9 IMPROVING PERFORMANCE Table 9.1: Basic Counters for Monitoring Windows Performance Object Counter Explanation Desired Values How to Improve Processor % Total Processor Time Total real work being performed by all system CPUs. Less than a 70% sustained value. Add a CPU or get a faster CPU. System Processor Queue Length The total number of threads in the queue waiting for a processor. If this value is consistently more than 2, this may indicate a CPU bottleneck. Add a CPU or get a faster CPU. Memory Pages/sec The number of 4KB pages written to or read from the paging file. Fewer than 5 15 pages per second. Sustained larger values can indicate a low available memory situation. Add more RAM or remove the demand for additional memory by reallocating resources. Memory Available Mbytes Amount of memory available after all process and caching memory have been allocated. If this number is consistently less than 15MB, consider adding RAM. This number should never drop to less than 4MB. Add more RAM. Paging File % Usage (Total instance) Total percentage of usage of the paging file. If this value is greater than 75%, the page file is being heavily used. Increase the initial size of the page file. Look at other memory-related counters to determine whether there is a memory bottleneck. Network Bytes Total/sec Total number of bytes transferred to the network per second. Should not approach the limits of the technology (such as 1.2MB/s for 10BaseT Ethernet or 12MB/s for 100BaseT). Move to a switched network technology or segment the network into smaller pieces. Physical Disk % Disk Time Percentage of time that a logical disk is used for both reads and writes. Should be less than 60% on a sustained basis. Add physical disks, faster disks, or faster disk controllers. Physical Disk Average Disk Queue Length The average number of pending read and write requests for a specific disk. Should be less than the number of spindles plus 2. Add additional physical disks, faster disks, or faster disk controllers.

Note: If you are looking for good and high quality web space to host and run your application check Lunarwebhost Low Cost Web Hosting services