AEM Cloud Services - BPO Report - Optimization Cache Hit Ratio

[Rohan_Garg]

Hi AEM Community,

I have received a BPO Report highlighting relevant technical KPIs for our AEM as a Cloud Service solution.

Please find the below extract from the report -

Optimization - This score provides a view of your optimisation against established AEM Cloud Service baseline metrics. It takes into consideration Cache Hit Ratio, and other performance and optimisation factors.

Cache Control -

In the Web First Paradigm, the performance of a service is determined by metrics such as Cache Hit Ratio and Cache Coverage. Adobe advises to reach a Cache Hit Ratio of 90% or even higher to see fast experiences in delivering the pages of the website.

From our httpderror logs we can see the cache hit ratio is averaging 70-80% as against Adobe's recommendation for 90%. However, from our logs we are seeing the below observations - 

1. A lot of requests which are either Adobe health check based and thus forwarded to publish instances without caching at Dispatcher.

For 16th Feb 2023, as per logs below counts were seen 

Requests with URI /system/probes/health - 197622 requests logged
Requests with URI /system/probes/live - 11472 requests logged
Requests with URI /system/probes/ready - 11480 requests logged

2. A lot of requests which seem to be bot/script originated (.php based) are blocked at Dispatcher (not at httpd).

The number of bot requests .php based requests blocked at Dispatcher - 1232 requests

Sample Endpoints - /wxo.php, /wp-info.php/, /hehe.php, /3x.php/?3x=3x and more

Query 1 - Are these requests which are blocked at dispatcher or the ones which are health check based requests responsible for skewing the cache hit ratio percentages ?

Query 2 - How can we improve these percentages for cache hit ratio regardless ?

As per Adobe's recommendations, we should ensure setting cache control header and expires header on every page. We have already done that and requested them to provide specific pages if they can find any.

@Jörg_Hoh, @arunpatidar, @SureshDhulipudi, @kautuk_sahni, @markus_bulla_adobe 

Thanks,

Rohan Garg

[arunpatidar]

Hi @Rohan_Garg 

Can you please ask adobe from where this data is collected?, Is it at CDN or dispatcher.

As you mentioned you have blocked bot request but those are blocked at httpd and CDN still considered those as request.

[Rohan_Garg]

Thanks for the reply @arunpatidar, I will ask Adobe this and get back to you!

[Rohan_Garg · Accepted solution]

@arunpatidar - The metric is from CDN in BPO for AEMaaCS implementation.

I have requested Adobe to reconsider their calculation for cache hit ratio when requests are bot based or service health based.

Will update their reply! Thanks for the support!