> I am curios at what point the the cache exceeds the comfort zone of the design.

In my opinion it depends more on the aspect how important is your cache / how quickly can you replace and repopulate it (how fast or loaded are your backends) / can your service work without it - as in if you have a single massive cache server what happens when it goes down?

I'm not sure what the upper or "reasonable" limits are but from a personal/practical experience I'm currently running instances with 32Gb ram and 4 x 1Tb ssds for cache files/zones (distributed via split_clients directive) each zone has keys_zone=2000m / each drive is filled up to 800G with roughly ~10M objects so in total 40-50M objects per server instance. Traffic Is arroung 1-2Gbits at peaks and there are no performance issues.

rr

_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx