Distributed Cache Read Misses counter

[therapynotes]

When looking at the Distributed Cache Read Misses counter what is considered a good/bad number? If we are over a bad threshold what should we do to correct it? Right now, we are getting ~44 read misses sampled every minute.

 

[admin]

The acceptable rate of cache misses depends on the requirements and design of your application and how the ScaleOut APIs are being used.

If your developers expect 100% of the ScaleOut API calls to successfully find an object in the ScaleOut service then 44 misses/sec could be interpreted as a problem. On the other hand, some applications use the ScaleOut service as a simple look-aside cache, where if an object resides in ScaleOut then great, it'll be used, otherwise it's no big deal--the value will be pulled from a database instead.

The overall request rate is worth considering, too: If you're doing 10,000 requests/sec and you only have 44 misses/sec then that's a good hit rate!