On Mon, 22 Aug 2016 15:45:52 +0200 Florent B wrote: > On 08/22/2016 02:48 PM, Christian Balzer wrote: > > Hello, > > > > On Mon, 22 Aug 2016 14:33:51 +0200 Florent B wrote: > > > >> Hi, > >> > >> I'm looking for informations about cache-tiering. > >> > > Have you searched the ML archives, including my > > "Cache tier operation clarifications" thread? > > I will read it thank you :) > > > > >> As I understand documentation, data written to hot storage is > >> immediately written to cold storage, so size=1 for hot storage should be > >> fine. > >> > > Where do you get that idea from? > > Please cite the source. > > If this were the case, where would be the speed advantage over your > > default pool? > > "immediately" is not the good word, but documentation says : "Ceph > clients write data to the cache tier and receive an ACK from the cache > tier. In time, the data written to the cache tier migrates to the > storage tier and gets flushed from the cache tier." > "In time" denotes "quickly", isn't it ? > Re-read that whole documentation again. In time may mean never, if the object stays hot and in the cache tier forever. You could configure things to FORCE all cache objects to be flushed out immediately or after a certain time, but that would be wasteful in the first case (same speed as the base pool) and pointless in the second one (the object may change immediately after it has been flushed and the base pool object copy would be worthless). And even if there would be just a short (lets say 5 seconds) delay between writes to the cache pool and flushes to the base pool, would you be comfortable with loosing all that data? Christian > > > > Journaling kinda resembles this behavior, cache tiering most definitely > > not. > > > > A cache pool is a regular Ceph pool, so it needs: > > a) journals (with filestore, for the time being) > > b) to be replicated at level you feel comfortable with. > > > > Whether b) is fine with 2 or should be 3 depends on the SSDs/NVMEs your > > using, the monitoring you're doing and of course your budget. > > > > A replication of 2 also will be faster than 3, but given that most if > > not ALL your hot objects will be in your cache pool and may NEVER be > > written to cold storage ever that shouldn't be your primary concern. > > > > Christian > > > >> Thank you. > >> > >> Florent > >> > >> _______________________________________________ > >> ceph-users mailing list > >> ceph-users@xxxxxxxxxxxxxx > >> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > >> > > > > > -- Christian Balzer Network/Systems Engineer chibi@xxxxxxx Global OnLine Japan/Rakuten Communications http://www.gol.com/ _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |