There's a nice whitepaper about under-provisioning everyone using SSDs should read it
http://www.sandisk.com/assets/docs/WP004_OverProvisioning_WhyHow_FINAL.pdf <http://www.sandisk.com/assets/docs/WP004_OverProvisioning_WhyHow_FINAL.pdf> But in reality I'd stick to the rated TBW/DWPD instead of guessing what the effect will be. And in reality the drives either die young (manufacturing defect), randomly or all at once (firmware error) and if they survive those types of defects they are likely to go well beyond their stated TBW. Some workloads are however much harder on the SSDs, especially those that have volatile cache. For example writing a 1B blocks synchronously will drain the reserves really fast while "host bytes written" remains extremely low. Jan > On 25 Aug 2015, at 17:02, Дробышевский, Владимир <[email protected]> wrote: > > Alexander, > > Samsung says that pm863 uses Samsung's "proprietary" 3-bit MLC V-NAND, and > sm863 - 2-bit MLC V-NAND. But from endurance perspective these drives don't > look as worth their money (except of some read-mostly cases). > > > С уважением, > Дробышевский Владимир > Компания "АйТи Город" > +7 343 2222192 > ICQ# - 1921011 > > Аппаратное и программное обеспечение > IBM, Microsoft, Eset > Поставка проектов "под ключ" > Аутсорсинг ИТ-услуг > > 2015-08-25 18:35 GMT+05:00 Alexandre DERUMIER <[email protected] > <mailto:[email protected]>>: > Hi, > > If it was me, I'll just waiting a little to see real endurance of theses > samsung drives, > because AFAIK, they use TLC memory (and I remember of all theses EVO drives > TLC memory bugs) > > But it seem that Dell use them is these new full flash array, > http://www.tomsitpro.com/articles/dell-ssd-3d-tlc-afa,1-2754.html > <http://www.tomsitpro.com/articles/dell-ssd-3d-tlc-afa,1-2754.html>, > so maybe they are quite ok. > > > > ----- Mail original ----- > De: "aderumier" <[email protected] <mailto:[email protected]>> > À: "Дробышевский, Владимир" <[email protected] <mailto:[email protected]>> > Cc: "Jan Schermer" <[email protected] <mailto:[email protected]>>, "ceph-users" > <[email protected] <mailto:[email protected]>> > Envoyé: Mardi 25 Août 2015 15:33:37 > Objet: Re: [ceph-users] Samsung pm863 / sm863 SSD info request > > Alexandre Derumier > Ingénieur système et stockage > > > Fixe : 03 20 68 90 88 > Fax : 03 20 68 90 81 > > > 45 Bvd du Général Leclerc 59100 Roubaix > 12 rue Marivaux 75002 Paris > > > MonSiteEstLent.com - Blog dédié à la webperformance et la gestion de pics de > trafic > > ----- Mail original ----- > De: "Дробышевский, Владимир" <[email protected] <mailto:[email protected]>> > À: "Jan Schermer" <[email protected] <mailto:[email protected]>> > Cc: "ceph-users" <[email protected] <mailto:[email protected]>> > Envoyé: Mardi 25 Août 2015 15:26:02 > Objet: Re: [ceph-users] Samsung pm863 / sm863 SSD info request > > Jan, thanks for the opinion! > > Do you think that 5.11 DWPD is terrible? For example, Intel S3610 has 3 DWPD > or 5475 TBW (but for 5 years, of course). > For 1.6TB Intel S3610 it means ~$0.175 per TB written. For pm863 3.84TB the > price per TB written will be ~$0.108. And more than twice capacity per drive > bay ratio. > > For sm863 1.84TB it's even better: $0.057 per TB written. Actually, it's > better than for Intel S3710 1.2TB: $0.081 per TB written for Intel. But s3710 > has better speed specs of course (the only problem that we have not enough > CPU speed for such IOPS, I think). > > I found Intel prices on http://google.com/shopping > <http://google.com/shopping> (the lowest ones) and the only one price I was > able to find for Samsung - through "Where to Buy" on the Samsung website. > For Intel prices were: > s3610 1.6TB - $1530 ($0.96 per GB of capacity) > s3710 1.2TB - $1780 ($1.49 per GB) > > For Samsung: > pm863 3.84TB - $2328 ($0.61 per GB) > sm863 1.92TB - $1324 ($0.69 per GB) > > As far as I know these are retail products (they are listed on Samsung > Business SSD section) but they aren't avaliable in my country and Support > said nothing about timescales. > > But I belive these SSDs will be avaliable in Europe and USA soon so I > consider them as an alternative to Intel drives for the new cluster I'm > working on. The only question is real speed. > > Best regards, > Vladimir Drobyshevskiy > > 2015-08-25 16:56 GMT+05:00 Jan Schermer < [email protected] > <mailto:[email protected]> > : > > > pm863 has terrible DWPD, I wouldn't even consider it. > sm863 is better and could be worthwile > > But I think both of these are OEM-only. Or can you buy them in bulk somewhere? > > Jan > > > > On 25 Aug 2015, at 12:18, Дробышевский, Владимир < [email protected] > > <mailto:[email protected]> > wrote: > > > > Hello! > > > > For a few months I'm looking for any news about new VNAND Samsung sm863 \ > > pm863 SSDs. A couple of weeks ago it's finally appeared on the Samsung > > web-site. Final specs look very interesting: > > > > sm863: up to 1.92GB, 520 / 485 mb/s sequential read / write, 97k / 29k > > random IOPS, cas latency 100, warranty: 5 years and 12320 TBW, suggested > > retail price $1260 (1.92TB model) > > > > pm863: up to 3.84GB, 540 / 480 mb/s sequential read / write, 99k / 18k > > random IOPS, cas latency 170, warranty: 3 years and 5600 TBW, suggested > > retail price $2200 (3.84TB model) > > > > So I would like to ask: has anybody got it yet and did any tests (ceph > > tests preferably)? > > > > Thanks! > > > > Best regards, > > Vladimir Drobyshevskiy > > _______________________________________________ > > ceph-users mailing list > > [email protected] <mailto:[email protected]> > > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > <http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com> > > > > > > > _______________________________________________ > ceph-users mailing list > [email protected] <mailto:[email protected]> > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > <http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com> >
_______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
