How To Guides > Application Performance Scores (APS) > Application Performance Report

Exinda Solution Center: Application Performance

The application performance solution monitors your important applications. On one screen, you can have insight into who is using the application and their perceived performance of the application. You can also see the amount of bandwidth used by the application and the amount of reduction achieved, if applicable. These charts can answer questions such as:

The application performance chart shows the network user experience of the performance of the application. You should expect the application performance to show a good score (between 8.5 and 10.0). If the score is less than 7.0, you may want to investigate if there is a problem. To see the measures that contributed to the score, press the ‘Show more’ link.

When the Application Performance monitor is first created, the application performance metric needs a baseline of observed traffic for the application in your network. The baseline operation will be started automatically and will observe an hour of traffic. Once a baseline is set, it will compare current traffic against the metrics calculated in the baseline period to determine the application performance score.

The inbound and outbound bandwidth charts show how much bandwidth the application is using. You should expect the bandwidth to show spikes instead of raised flat tops. Flat tops in the graph often indicate that the traffic may be limited by policy rules. The chart shows data measured on the WAN-side of the appliance, that is before the accelerated traffic is decompressed for inbound traffic and after acceleration and traffic shaping policies have been applied for outbound traffic. You can choose to overlay the data measured on the LAN-side of the appliance. This will show the amount of reduction achieved due to acceleration and traffic shaping.

The users and hosts bar charts show the WAN-side data volumes consumed by the top users and hosts for the application. Typically, applications are used by multiple users or hosts and the traffic distribution is fairly even amongst the top users or hosts. If one user or host shows considerably more data volume than the other users, it may be reasonable behavior or it may indicate a problem worthy of further investigation. Also, you can choose to show just internal endpoints, that is, hosts and users on the LAN-side of your appliance, or just external endpoints, that is, hosts and users on the WAN-side of your appliance. You can also choose to show just users, just hosts, or both.

ClosedIf the application performance chart is baselining

The application performance score is calculated as the result of analyzing network performance of the application relative to thresholds set for various network and server metrics. When the application performance object is first set up, the baseline thresholds are set by observing traffic for a period of time. If the application performance chart indicates that it is baselining, it is computing the initial threshold values. Once the baseline period is complete, the application performance chart will show the results. When the Application Performance report is first created, it will automatically start to baseline the traffic. If no traffic is observed for the specified application during the baselining period, a new baseline will be repeatedly started until traffic is observed and thresholds are calculated.

ClosedIt said it was going to baseline for an hour but it completed in 10 minutes. Was it a valid baseline period?

Yes, it is a valid period. The Exinda appliance is always observing traffic and it stores its observations for an hour. If near the end of an hour (for instance, at 9:50) you create an application performance page, then the application performance score object will start baselining and can look at the last hour of data collected for the application and so can complete the baseline period in the remaining 10 minutes of the hour.

ClosedTo set a new baseline for application performance thresholds

Go to Configuration > Objects > Service Levels > Application Performance Score (APS) and edit the APS object. It will be named the application name prefixed with “Solution Center” and postfixed with an application ID. Select the Auto Baseline Period and press the Start Baseline button. It is recommended to baseline the application when you expect the application to perform reasonably well.

ClosedTo investigate a poor application performance score (APS)

Click on the Show more link beside the application performance chart. A new screen charts the measures contributing to the APS:

  • Network delay and normalized network delay – the amount of time for data to traverse the network (on the wire)
  • Server delay and normalized server delay – the amount of time for a server to respond to a request
  • Round trip time – the amount of time for data to travel from a device across a network and return
  • Jitter – a measure of the variability of Network Delay. We define it as one standard deviation of the Network Delay.
  • Inbound loss and outbound loss – the amount of data retransmitted

Inspect the charts to determine which attribute caused the poor APS score. For example, if the server delay measures are good but the network delay measures are bad, then you know that the network is to blame and perhaps you can do something about it. If the network delay measures are good but the server delay measures are bad, then you should have someone investigate why the server is performing poorly.

Note that if the baselining period was not typical, then the calculated thresholds may be overly high or low. For example, if the application was baselined on a weekend when there was very little traffic, then the thresholds may be much lower than would be expected when the network is in a typical use scenario. Similarly, if the application was baselined during an extremely busy time, such as when most employees are watching an online CEO webcast, then the thresholds may be much higher than would be expected when the network is in a typical use scenario.

ClosedTo investigate unusual throughput behavior for this application, such as raised flat tops

If the flat tops have a corresponding drop in application performance score, you likely have an issue that you can address by policy changes. That is, you can ensure that the policy (or policies) that control this application have adequate guaranteed bandwidth or that other lesser important apps, such as recreational apps, have a lower maximum bandwidth.

ClosedTo determine throughput values for specific points in time in the throughput chart

Hover your cursor over the chart. A data brush will appear showing average throughput for the specific point in time.

ClosedTo investigate reduction due to acceleration, edge cache, and bandwidth shaping

Toggle on the LAN-side reporting to overlay the LAN-side values on the WAN-side values. The difference between the LAN-side and WAN-side throughput lines indicates the amount of reduction that was achieved. Note that the total reduction in data volume as a percentage is shown under the chart.

ClosedTo investigate application usage by the top internal users and top internal hosts

Ensure the Internal button is toggled on and the External button is toggled off. The top hosts and top users (if configured) found on the LAN-side of the appliance are shown.

ClosedTo investigate application usage by the top external users and top external hosts

Ensure the External button is toggled on and the Internal button is toggled off. The top external hosts and top external users (if configured) found on the WAN-side of the appliance are shown.

| Exinda.com

Using the help

Last Updated: Monday, February 1, 2016 at 12:24 PM

Exinda Management Center 2.0 ExOS 7.4.1 | Copyright © 2016 Exinda Networks, Inc.