I really like the 'Percentage of TCP sessions with errors' as opposed to the count of the number of errors because it allows you to quickly relate to the business impact it could be having on the end user.
For application level errors (HTTP, Citrix, SAP) is there a similar percentage metric that could be used?
We have to be careful and take protocol by protocol. Most of the errors and availability will fall on this three kinds of errors (TCP, Transport and Application)
Almost all the application protocols fall in this classification. Then For the HTTP:
- TCP: Network errors
- Transport: HTTP Errors
-Availability (Application): The custom defined errors
For most of the application protocols works almost(some of them has defined application errors) like this (LDAP,RMI,etc.)
For the SAP we need to be careful because most of the related errors on SAP are not transports errors are errors that fails on the application (We have to notice that SAP is an application it self and not a application protocol). Then in this case most of the errors fall in the application side
For Citrix. Citrix is a middleware that make the network transport between. For the moment on Citrix the errors Falls on the TCP, let see on the future.
Let me know if you want more detail on other protocols.
I hope it helps