Monitoring performance

In this topic

• Viewing server statistics as charts
• Request totals
• Request throughput
• Timeout rate
• Error rate
• Request profile
• Wait time
• Utilization

You can use Manager to view charts that describe your server's performance. A chart puts information from the log files into an easy-to-read graphical format. The information in a chart can help in your proactive monitoring as an administrator, since charts make it easier to visualize server usage patterns that may lead to problems.

The Manage my GIS Server link contains a link called Charts. On the Charts tab, you can view predefined charts, or customize the charts to display only the information you want to see. The types of charts available for viewing are described here.

Request totals categorizes all requests to the server as either successful, timeouts, or errors, and displays the resulting information. These charts show the load on the server in terms of the number of requests processed by the server as a whole, by host machines, and by server objects. You can see which server objects are the most popular in terms of the total number of requests, and which ones are experiencing the most problems in the form of errors and timeouts. If problems are occurring on a specific host machine, it will be apparent from these charts.

Request throughput shows the total number of successful requests to the server in a given time period. If two machines of comparable power experience different levels of throughput, one of the machines may have a problem. The throughput charts show you what time of day your server objects are used the most. This information will help you determine appropriate recycling times for server objects.

You can also view the range of throughput for your server objects. If a server object has a wide range of throughput, you should set a wide range for the minimum and maximum number of instances. However, if the load on a server object is somewhat steady, you can set a narrow range between the minimum and maximum number of instances.

Throughput measures the number of times a context is handed out and released; therefore, throughput for pooled objects will generally be higher than throughput for non-pooled objects, even if the non-pooled objects experience heavy use.

Timeout rate displays statistics for both usage and wait timeouts. Usage timeouts occur when a client uses a server object beyond the maximum usage time. Wait timeouts occur when a client waits to get a server object beyond the maximum wait time. You can use the timeout rate charts to get an overview of the number of timeouts that are occurring for the GIS server, for individual host machines, and for server objects.

If a server object is experiencing excessive usage timeouts, it may mean that the server object is consistently having a problem completing a certain task. If this is the case, check your server object and its associated data to ensure they are configured correctly. If the server object is working as it should, you can increase the maximum usage time for the server object.

An increased number of wait timeouts indicates a rise in demand for a server object. In this situation, consider creating more instances of the server object or increasing the maximum wait time.

Error rate shows the number of errors that the server tracks. The server records two kinds of errors: server object creation errors and server context creation errors—such as when a client requests a server context but access is denied. If a particular machine is experiencing high error rates, it may have a hardware or configuration problem.

Request profile is a representation of what happens to a request for a server context, including wait time, creation time, and usage time.

One of the more interesting profiling charts shows how much time a client has to wait to get a server context, because this is a measure of the speed of a client application. Wait time is a combination of the time a client spends waiting in a queue and the time the server takes to create the server object. If the average wait time for a server object approaches the object's maximum wait time, consider increasing the maximum wait time or creating more instances of the object.

Utilization measures the average number of instances of a server object in use during a given time period. Utilization statistics for the GIS server and individual host machines are derived by adding the utilization statistics of all the server objects in the configuration.

You can use the utilization charts to see the average number of server object instances for your GIS server and compare it to the capacity of the machines that comprise the server. If the number of instances is close to the capacity, it may be time to add a new machine.

You can also compare the average number of instances running on each of your host machines. If all host machines are equivalently configured, the average number of instances should be about the same across all machines, indicating that the workload is balanced.

Perhaps the most interesting feature of the charts is the ability to view highest and lowest server object utilization. You can use this information to determine the number of server object instances to make available. For example, if you have a server object that is highly utilized, consider increasing the available instances of that server object.

Although you can view utilization for a variety of time periods, longer time periods yield more accurate results. This is because usage times, especially for a non-pooled object, can span beyond the charted time period. The server records usage time when a client finishes using a server object. This introduces into the chart a potential for error, which is minimized when you chart a longer time period.