Can Critical Performance Affect Availability?

govitallen

I am trying to improve my availability metric accuracy for situations where everything appears to be working but the performance is effectively a system down.  For example a DDOS attack may effectively take down a system as far as public access, but from inside the network everything appears fine.  Traffic may be getting queued to the point that it appears that nothing is getting through.  There is no apparent outage but it is effectively down.  There are performance counters and other statistics we can poll that indicate a "Critical" condition but our availability is still 100%. 

How can I use the critical performance to reflect the actual availability (or rather unavailability) of the system? 

jm_sysadmin

So you definitely do want to move beyond simple up down once you have that taken care of. I will say that for every app its going to look a little different, but the big check is the response time of the pages that you are checking, and maybe try to hit an API to pull data. Some of my systems we retrieve a batch of records, and track performance over time. Sometimes its 1 record sometimes its a 1000, it all depends on the system and where you expect to be able to see the slowness. 

Any synthetic check of what users do is helpful (think about the WPM module) or the DHCP/DNS checks that include response time. Scripts can open a lot of options. 

Another thing to watch for from a network perspective is dropped packets or other network related things. If you are just focused on SAM stuff, you can pull things like the windows counters below:
Network-Related Performance Counters | Microsoft Docs 

It tells you a bit about the environment right at the server, but personally, I look at the network devices, and we other systems looking at the packets themselves for meta data.