A poster in another forum mentioned that Seagate (and Hitachi, amongst others) is now selling something labeled as "NearLine SAS" storage (e.g. Seagate''s NL35 series). Is it me, or does this look like nothing more than their standard 7200-rpm enterprise drives with a SAS or FC interface instead of a SATA one? I can''t see any real advantage of those over the existing enterprise SATA drives (e.g. Seagate''s Constellation ES series), other than not needing a FC/SAS->SATA gateway in the external drive enclosure. -- Erik Trimble Java System Support Mailstop: usca22-123 Phone: x17195 Santa Clara, CA
On Tue, Jan 19, 2010 at 12:16 AM, Erik Trimble <Erik.Trimble at sun.com> wrote:> A poster in another forum mentioned that Seagate (and Hitachi, amongst > others) is now selling something labeled as "NearLine SAS" storage (e.g. > Seagate''s NL35 series). > > Is it me, or does this look like nothing more than their standard 7200-rpm > enterprise drives with a SAS or FC interface instead of a SATA one? > > I can''t see any real advantage of those over the existing enterprise SATA > drives (e.g. Seagate''s Constellation ES series), other than not needing a > FC/SAS->SATA gateway in the external drive enclosure. > > >Seagate claims the SAS versions of their drives actually see IOPS improvements: http://www.seagate.com/www/en-us/products/servers/barracuda_es/barracuda_es.2 If the SAS version is dual ported like I would expect, that''s also a MAJOR benefit. -- --Tim -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100119/0f6a3d40/attachment.html>
Tim Cook wrote:> > > On Tue, Jan 19, 2010 at 12:16 AM, Erik Trimble <Erik.Trimble at sun.com > <mailto:Erik.Trimble at sun.com>> wrote: > > A poster in another forum mentioned that Seagate (and Hitachi, > amongst others) is now selling something labeled as "NearLine SAS" > storage (e.g. Seagate''s NL35 series). > > Is it me, or does this look like nothing more than their standard > 7200-rpm enterprise drives with a SAS or FC interface instead of a > SATA one? > > I can''t see any real advantage of those over the existing > enterprise SATA drives (e.g. Seagate''s Constellation ES series), > other than not needing a FC/SAS->SATA gateway in the external > drive enclosure. > > > > Seagate claims the SAS versions of their drives actually see IOPS > improvements: > http://www.seagate.com/www/en-us/products/servers/barracuda_es/barracuda_es.2 > > If the SAS version is dual ported like I would expect, that''s also a > MAJOR benefit. > > -- > --Timstupid question here: I understand the advantages of dual-porting a drive with a FC interface, but for SAS, exactly what are the advantages other than being able to read and write simultaneously (obviously, only from the on-drive cache). And yeah, these Seagates are dual-ported SAS. (according to the spec sheet) Also, a 38% increase in IOPS without LESS drive cache seems unlikely. Or, at least highly workload-dependent. Check that, they''re claiming 38% better IOPS/watt over the SATA version, which, given that the SAS one pulls 10% more watts, means in absolute terms 45% or so. I''m really skeptical that only an interface change can do that. -- Erik Trimble Java System Support Mailstop: usca22-123 Phone: x17195 Santa Clara, CA
On Tue, Jan 19, 2010 at 1:06 AM, Erik Trimble <Erik.Trimble at sun.com> wrote:> stupid question here: I understand the advantages of dual-porting a drive > with a FC interface, but for SAS, exactly what are the advantages other than > being able to read and write simultaneously (obviously, only from the > on-drive cache). > And yeah, these Seagates are dual-ported SAS. (according to the spec sheet) >Path redundancy. While it''s fairly rare, paths to drives do go down. Redundancy is a good thing :)> Also, a 38% increase in IOPS without LESS drive cache seems unlikely. Or, > at least highly workload-dependent. > Check that, they''re claiming 38% better IOPS/watt over the SATA version, > which, given that the SAS one pulls 10% more watts, means in absolute terms > 45% or so. I''m really skeptical that only an interface change can do that. > >Without benchmarking myself, I can''t really speak much to their claims. I WILL however say it''s VERY unlikely they''d drop the cache on something intended for the enterprise without being extremely confident its performance would be the same or better. It wouldn''t surprise me at all to hear the components they use for their SAS interfaces yield significantly better performance. Plus, if it''s dual ported... I wouldn''t expect to see 38% consistently, but I would expect to see better performance across the board. -- --Tim -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100119/4f064dbc/attachment.html>
> A poster in another forum mentioned that Seagate (and Hitachi, amongst > others) is now selling something labeled as "NearLine SAS" storage > (e.g. Seagate''s NL35 series).Industry has moved again. Better get used to it. Nearline SAS is a replacement for SATA. It''s a lower cost drive than SAS, with higher reliability than SATA. I have begun seeing vendors that sell only SAS and NearLine SAS. (Some dell servers.) SATA is not dead. But this will certainly change things up a bit.
Edward Ned Harvey wrote:>> A poster in another forum mentioned that Seagate (and Hitachi, amongst >> others) is now selling something labeled as "NearLine SAS" storage >> (e.g. Seagate''s NL35 series). > > Industry has moved again. Better get used to it. > > Nearline SAS is a replacement for SATA. It''s a lower cost drive than SAS, > with higher reliability than SATA. I have begun seeing vendors that sell > only SAS and NearLine SAS. (Some dell servers.) > > SATA is not dead. But this will certainly change things up a bit.Actually, this sounds like really good news for ZFS. ZFS (or rather, Solaris) can make good use of the multi-pathing capability, only previously available on high speed drives. Of course, ZFS can make use of any additional IOPs to be had, again only previously available on high speed drives. However, ZFS generally doesn''t need high speed drives and does much better with hybrid storage pools. So these drives sound to me to have been designed specifically for ZFS! It''s hard to imagine any other filesystem which can exploit them so completely. -- Andrew
On Tue, Jan 19, 2010 at 4:17 AM, Andrew Gabriel <Andrew.Gabriel at sun.com>wrote:> > Actually, this sounds like really good news for ZFS. > ZFS (or rather, Solaris) can make good use of the multi-pathing capability, > only previously available on high speed drives. > Of course, ZFS can make use of any additional IOPs to be had, again only > previously available on high speed drives. > However, ZFS generally doesn''t need high speed drives and does much better > with hybrid storage pools. > > So these drives sound to me to have been designed specifically for ZFS! > It''s hard to imagine any other filesystem which can exploit them so > completely. > > >You mean like WAFL? -- --Tim -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100119/dc29f8e5/attachment.html>
I was today reasearching this same phenomenon. The multipath is required for HA storage solutions with redundant i/o path backplanes and redundant controllers. ( If a controller fails, the other one can still access the harddisk.) I read about an LSI SAS-to-SATA bridge what can be attacched onto an ordinary SATA drive to make it operate like a SAS drive (e.g. be able to multi path to that drive.) Is there anyone on the list that can give some information about this? I am google-ing my pants of to find some kind of shop selling these bridges. I want to build a redundant JBOD to share between two ZFS hosts.(Active/Standby) I already found the nearline SAS disks, but this will not fix the multipath problem for SATA SSD disks. Regards, Armand ----- Original Message ----- From: "Erik Trimble" <Erik.Trimble at Sun.COM> To: "Tim Cook" <tim at cook.ms> Cc: "zfs-discuss" <zfs-discuss at opensolaris.org> Sent: Tuesday, January 19, 2010 8:06 AM Subject: Re: [zfs-discuss] "NearLine SAS"?> Tim Cook wrote: >> >> >> On Tue, Jan 19, 2010 at 12:16 AM, Erik Trimble <Erik.Trimble at sun.com >> <mailto:Erik.Trimble at sun.com>> wrote: >> >> A poster in another forum mentioned that Seagate (and Hitachi, >> amongst others) is now selling something labeled as "NearLine SAS" >> storage (e.g. Seagate''s NL35 series). >> >> Is it me, or does this look like nothing more than their standard >> 7200-rpm enterprise drives with a SAS or FC interface instead of a >> SATA one? >> >> I can''t see any real advantage of those over the existing >> enterprise SATA drives (e.g. Seagate''s Constellation ES series), >> other than not needing a FC/SAS->SATA gateway in the external >> drive enclosure. >> >> >> >> Seagate claims the SAS versions of their drives actually see IOPS >> improvements: >> http://www.seagate.com/www/en-us/products/servers/barracuda_es/barracuda_es.2 >> >> If the SAS version is dual ported like I would expect, that''s also a >> MAJOR benefit. >> >> -- >> --Tim > stupid question here: I understand the advantages of dual-porting a drive > with a FC interface, but for SAS, exactly what are the advantages other > than being able to read and write simultaneously (obviously, only from the > on-drive cache). > And yeah, these Seagates are dual-ported SAS. (according to the spec > sheet) > > Also, a 38% increase in IOPS without LESS drive cache seems unlikely. Or, > at least highly workload-dependent. > Check that, they''re claiming 38% better IOPS/watt over the SATA version, > which, given that the SAS one pulls 10% more watts, means in absolute > terms 45% or so. I''m really skeptical that only an interface change can > do that. > > > -- > Erik Trimble > Java System Support > Mailstop: usca22-123 > Phone: x17195 > Santa Clara, CA > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
On Tue, Jan 19, 2010 01:02 PM, "A. Krijgsman" <a.krijgsman@draftsman.nl> wrote:> I was today reasearching this same phenomenon. > The multipath is required for HA storage solutions with redundant i/o path > backplanes and redundant controllers. > ( If a controller fails, the other one can still access the harddisk.) > > I read about an LSI SAS-to-SATA bridge what can be attacched onto an > ordinary SATA > drive to make it operate like a SAS drive (e.g. be able to multi path to > that drive.) > Is there anyone on the list that can give some information about this? > > I am google-ing my pants of to find some kind of shop selling these bridges. > I want to build a redundant JBOD to share between two ZFS > hosts.(Active/Standby) > > I already found the nearline SAS disks, but this will not fix the multipath > problem for SATA SSD disks. > > Regards, > ArmandI assume you mean these type of interposer cards? e.g http://www.lsi.com/DistributionSystem/AssetDocument/SCG_LSISS9252_PB_082709.pdf Based on my experiences, they generally go to VARs or OEMs e.g. Dell, HP. I do recall seeing some older gens on a 2950 that passed my way..
On Tue, Jan 19, 2010 01:36 PM, "A. Krijgsman" <a.krijgsman@draftsman.nl> wrote:> Yes, those i meant! (interposer cards, could get on that name!) > The Dell''s and HP''s would al work in any enviroment? > > It''s just plain converting between industry standards right? > > Or do people have a better solution to share a single SSD from a JBOD > between > two hosts? > > Regards, > Armand >Perhaps I didn''t read your post in it''s entirety. If your goal is to share a single drive between two jbod systems, an interposer card won''t help. There was some talk on the list about sharing SSDs over low latency networks.