Is it the server?
In Is it the network? we used the Network Performance dashboard to take a close look at response time issues from a network perspective. Now we will focus on response time issues from a server perspective.
The Server Performance dashboard helps you determine if a server is impacting server slow and is having slow response times that are affecting users.
Select the Server Performance dashboard. To get the dashboard, see Importing a user-contributed dashboard.
Slow server problems are typically caused by:
Many users accessing an application server at the same time.
The server not having enough resources (CPU, memory, etc).
The server is encountering error conditions.
Dependent front end or middleware, database, server or other systems are causing delays.
 
This dashboard provides detailed server performance metrics as well as some metrics we previously examined for the network. Since network performance impacts server response times, it’s convenient to have analytics from both in one dashboard. For Figure 53, we have filtered the dashboard to a single HTTP server.
 
Figure 53: Server Performance dashboard
 
Looking at the Servers by response time max widget, we see the maximum response time is greater than four seconds. But just as important, notice that the average response time is only 159 milliseconds and the round trip delay is very low (40 milliseconds). This means that the majority of the time the server response time is low and our network is running efficiently, but we do have some performance spikes.
Let’s look at another widget in Figure 53: Server/client by response time max. These are the clients connecting to our HTTP server sorted by maximum response time. You can see that only one client has a high maximum response time value (4137 milliseconds), the rest are low. Also notice that the server had 536 retransmissions to that client. This suggests that the problem is limited to this server/client pair.
Tip! Using additional tools like Connection Dynamics or Ad Hoc Application Transaction Analysis in Observer to look at the packet data between this client and server can provide detailed insight into a problem such as this.
 
Let’s look at another example: Figure 54.
High average response time and low network delay is indicative of a slow server problem. The two servers not only have a high maximum response time, but their average response time is greater than one second. We know that the network round trip delay is relatively low so the problem is not the network.
 
Figure 54: Server Response Time - Max
 
A correlation chart can show if there is a relationship between the number of connections, traffic volumes, network delay, or retransmissions and the application server response time.
 
Figure 55: Correlation chart: performance analysis by time
 
You can also use the Server Analysis tool in Observer to get server details. In Figure 56, you can see values comparing the number of users and simultaneous requests. Notice that the response time is relatively linear and flat, indicating the server is handling the load well. If you saw a large or exponential increase in response time as simultaneous requests increased, that would suggest too many users or inadequate server resources. We know that the average response time is relatively flat, so the problem is not the network.
 
Figure 56: Average response time