Re:Ristretto: A High Performance, Concurrent, Memory-Bound G...
by wrnr on Saturday September 21, 2019 @02:37PM
(#21037071)
Very much to the contrary, why go over the network to get a value from a key-value store when you can have a mutex protected map that only hods the values that where accessed by that node. The "lets have a shared memory between our services" architecture makes things easy only to make it much harder again.
[ Parent
| on HN ]
Re:Ristretto: A High Performance, Concurrent, Memory-Bound G...
by hinkley on Saturday September 21, 2019 @07:25PM
(#21038555)
I’ve been trying to convince my coworkers that we should have a cache per node, but so far they aren’t biting.
Alternatively I think within the decade we are going to expect every service to contain its own fault prevention logic instead of just the web tier. At which point every service has its own reverse caching proxy in front of it.