Question on the Current Cache Hit Ratio log messages in dispatcher logs

Avatar

Avatar

ToddRowan

Avatar

ToddRowan

ToddRowan

02-12-2020

I recently learned about the Current cache hit ratio log message in dispatcher logs. 

 

But I can't find any good documentation on it. So I have a few questions:

  1. Over what timeframe is the ratio calculated? Is it from Apache startup?
  2. Does the calculation ignore those paths that are configured to never cache in the dispatcher?

I ask because I have an existing access log query that calculates the difference between dispatcher requests and publish instance requests only on paths where I know there is caching. I ignore servlet paths that do no caching, and I also ignore all requests other than 200s. With that calculation I get a higher hit ratio than what the dispatcher claims. 

 

How much should I trust the dispatcher log calculation?

Replies

Avatar

Avatar

sureshd71454102

Avatar

sureshd71454102

sureshd71454102

03-12-2020

Cache Hit Ratio: This means your fetching from render level vs from cache is measured.  You want to hit 80+ percent from cache, and you should follow the help here:

https://helpx.adobe.com/experience-manager/kb/optimizing-the-dispatcher-cache.html

 

For common dispatcher log details:

https://helpx.adobe.com/experience-manager/kb/ams-dispatcher-manual/common-logs.html

 AMS Dispatcher Manual

https://helpx.adobe.com/experience-manager/kb/ams-dispatcher-manual.html

 

Avatar

Avatar

ToddRowan

Avatar

ToddRowan

ToddRowan

03-12-2020

Yes, I've seen all that. None of this answers any of my questions.

Avatar

Avatar

Jörg_Hoh

Employee

Total Posts

3.0K

Likes

953

Correct Reply

1.0K

Avatar

Jörg_Hoh

Employee

Total Posts

3.0K

Likes

953

Correct Reply

1.0K
Jörg_Hoh
Employee

03-12-2020

 

The ratio is calculated as (100 * req_hits / req_total) with

 

req_hits = # files delivered from cache

req_total = the total number of requests handled by the dispatcher

 

from this point of view the cache-hit ratio is the same if requests are blocked on dispatcher (not httpd!) or forwarded to publish. So the number is nice, but it's hard to draw meaningful conclusions from it.