On 07/22/2013 05:53 PM, Chen, Xiaoxi wrote: > Basically i think endurance is most important for a ceph journal，since the workload for journal is full write，you can easily caculate how long your ssd will burn out.. even we assume your ssd only run at 100MB/s in average，you will burn out 8TB/day and 240TB/month > Well, "only" 100MB/sec? I still have to see the first cluster that does 100MB/sec inbound traffic on EACH node 24/7. I've been using Intel X25-M 80GB SSDs in MySQL servers for over 3 years now and haven't seen a single one wearing out. Those servers (~25) are being used intensively with a dataset of about 50GB each. It will depend on the cluster you are using, but 100MB/sec sustained for 24 hours a day on each node is a very heavily used cluster, I really wonder if your backing disks are able to keep up. All the static tests to wear an SSD out are cool, but they are far from reality imho. You should also underpartition the SSD, so buy a 200GB SSD and only use 16GB in the SSD so the wear-leveling algorithm can do it's job. A 200GB MLC SSD (10.000 writes) will wear out in about 400 days if you are writing 60MB/sec to it 24/7, still a very high number imho. Wido > DCS 3500 is definitely not useable in this case unless you want to swap your ssd every single month.. > > > > 在 2013-7-22，22:47，"Mark Nelson" <mark.nelson@xxxxxxxxxxx> 写道： > >> On 07/22/2013 09:30 AM, Gandalf Corvotempesta wrote: >>> 2013/7/22 Mark Nelson <mark.nelson@xxxxxxxxxxx>: >>>> I don't have any in my test lab, but the DC S3700 continues to look like a >>>> good option and has a great reputation, but might be a bit pricey. From that >>>> article it looks like the Micron P400m might be worth looking at too, but >>>> seems to be a bit slower. >>> >>> DC S3500 should be the same (for a journal) but at lower price. >> >> At least based on this: >> >> http://www.hardcoreware.net/intel-dc-s3500-480gb-ssd-review/ >> >> The write speeds are lower at the same capacity and the endurance appears to be significantly lower since the S3500 isn't using the HET cells. >> >> Personally I'm still thinking the 200GB S3700 is the way to go, but I haven't looked too closely at the micron drive or other options out there. Something like the Marvell Dragonfly could be very interesting for larger nodes, but I haven't tested it yet. >> >> Mark >> _______________________________________________ >> ceph-users mailing list >> ceph-users@xxxxxxxxxxxxxx >> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > -- Wido den Hollander 42on B.V. Phone: +31 (0)20 700 9902 Skype: contact42on _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |