I've been using Mod_Pagespeed for many years now, and recently moved to REDIS for storing its optimized resources (using ModPagespeedRedisServer and ModPagespeedRedisDatabaseIndex). A regular disk cache is configured as well, as per instructions, though my understanding is that the bulk of cached resources will go to REDIS. It's working fine so far.
My question is about the need (or not) of persistence for the REDIS cache: We use sharding in Mod_Pagespeed (via ModPagespeedShardDomain) and Amazon Cloudfront as a CDN. Let's say I set up REDIS so it no longer saves itself to disk (ie. no persistence, it will start empty every time there is a restart). I understand Pagespeed will have to recreate the optimized resources each time it restarts and "warm up the cache" again. I'm fine with that. What I would like to know is if the optimized resources will get the same name, in other words, is their name based on the original file name, modification date and such with a hash function that creates the exact same file name each time as long as the resource didn't change?
If that is the case (the optimized resource name and URL doesn't change, even after a restart/recreation) it would mean the CDN does not have to re-fetch all the resources, things would work nicely. If on the other hand the resource names are random, and change with each time they are optimized, it would mean sending over multiple Gigabytes to the CDN with each restart, and that is something I would like to avoid.
I would like to set REDIS so it no longer uses a disk-based backend, and save a few Gb in storage (and unneeded disk writes), but not if it means sending all those optimized resources over to the CDN with every restart.
Anyone know the answer to this one?