It looks like Intel has a huge hit (product) on its hands with the
latest SSD product announcements. No pricing yet ... but the specs
will push computer system IO bandwidth performance to numbers only
possible today with extremely expensive RAM based disk subsystems.
SSDs + ZFS - a marriage made in (computer) heaven!
http://www.tomshardware.com/news/intel-idf-ssd,6205.html
http://www.theinquirer.net/gb/inquirer/news/2008/08/19/idf-2008-intel-ssd-cometh
Regards,
--
Al Hopper Logical Approach Inc,Plano,TX al at logical-approach.com
Voice: 972.379.2133 Timezone: US CDT
OpenSolaris Governing Board (OGB) Member - Apr 2005 to Mar 2007
http://www.opensolaris.org/os/community/ogb/ogb_2005-2007/
On Wed, 20 Aug 2008, Al Hopper wrote:> It looks like Intel has a huge hit (product) on its hands with the > latest SSD product announcements. No pricing yet ... but the specs > will push computer system IO bandwidth performance to numbers only > possible today with extremely expensive RAM based disk subsystems. > > SSDs + ZFS - a marriage made in (computer) heaven!Where''s the beef? I sense a lot of smoke and mirrors here, similar to Intel''s recent CPU "announcements" which don''t even reveal the number of cores. No prices and funny numbers that the writers of technical articles can''t seem to get straight. Obviously these are a significant improvement for laptop drives but how many laptop users have a need for 11,000 IOPs and 170MB/s? It seems to me that most laptops suffer from insufficent RAM and low-power components which don''t deliver much performance. The CPUs which come in laptops are not going to be able to process 170MB/s. What about the dual-ported SAS models for enterprise use? Bob =====================================Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
On Wed, Aug 20, 2008 at 5:17 PM, Bob Friesenhahn < bfriesen at simple.dallas.tx.us> wrote:> On Wed, 20 Aug 2008, Al Hopper wrote: > > > It looks like Intel has a huge hit (product) on its hands with the > > latest SSD product announcements. No pricing yet ... but the specs > > will push computer system IO bandwidth performance to numbers only > > possible today with extremely expensive RAM based disk subsystems. > > > > SSDs + ZFS - a marriage made in (computer) heaven! > > Where''s the beef? > > I sense a lot of smoke and mirrors here, similar to Intel''s recent CPU > "announcements" which don''t even reveal the number of cores. No > prices and funny numbers that the writers of technical articles can''t > seem to get straight. > > Obviously these are a significant improvement for laptop drives but > how many laptop users have a need for 11,000 IOPs and 170MB/s? It > seems to me that most laptops suffer from insufficent RAM and > low-power components which don''t deliver much performance. The CPUs > which come in laptops are not going to be able to process 170MB/s. > > What about the dual-ported SAS models for enterprise use? > > Bob > =====================================> Bob Friesenhahn > bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ > GraphicsMagick Maintainer, http://www.GraphicsMagick.org/ > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discussI don''t know about that. I just went from an SSD back to a SATA drive because the SSD started failing in less than a month (I''m having troubles believing this great write-leveling they talk about is working properly...). And the SATA drive is dog slow in comparison. The biggest issue is seek times. Opening apps/directories there is a VERY noticeable difference from the SSD to this drive. The user experience is drastically improved with the SSD imho. Of course, the fact that it started giving me i/o errors after just 3 weeks means it''s going to be RMA''d and won''t find a home back in my laptop anytime soon. This was one of the 64GB OCZ Core drives for reference. --Tim -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20080820/ec35f800/attachment.html>
On Wed, 20 Aug 2008, Tim wrote:> > I don''t know about that. I just went from an SSD back to a SATA drive > because the SSD started failing in less than a month (I''m having troubles > believing this great write-leveling they talk about is working > properly...). And the SATA drive is dog slow in comparison. The biggest > issue is seek times. Opening apps/directories there is a VERY noticeable > difference from the SSD to this drive.The fact of the matter is that these new SSD drives are only 32GB or 80GB in size and will become available when 2TB hard drives become available. The 2TB hard drives will offer similar throughput but with far less IOPS. The SSD drives will work well for a boot drive, or a non-volatile transaction cache, but will be dramatically more expensive for storage than traditional hard drives. This must be why Intel is focusing on laptop users and not on enterprise storage. In spite of many vendor claims of reliability, most enterprise users are going to want to see these products deployed for a few years before they entrust them with their critical data. Bob =====================================Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
Bob Friesenhahn writes:> > The SSD drives will work well for a boot drive, or a non-volatile > transaction cache, but will be dramatically more expensive for storage > than traditional hard drives. This must be why Intel is focusing on > laptop users and not on enterprise storage. >The sweet spot will be non-volatile transaction cache coupled with slow, low cost and low power big hard drives. As non-volatile cache devices get bigger, the performance demands on the bulk storage become less. Ian.
Bob Friesenhahn wrote:>> >> SSDs + ZFS - a marriage made in (computer) heaven! > > Where''s the beef? > > I sense a lot of smoke and mirrors here, similar to Intel''s recent CPU > "announcements" which don''t even reveal the number of cores. No > prices and funny numbers that the writers of technical articles can''t > seem to get straight. > > Obviously these are a significant improvement for laptop drives but > how many laptop users have a need for 11,000 IOPs and 170MB/s? It > seems to me that most laptops suffer from insufficent RAM and > low-power components which don''t deliver much performance. The CPUs > which come in laptops are not going to be able to process 170MB/s.I guess you have not used current day laptops. I''ve used several brands that come standard with dual-core processors, 4 gig RAM, and 250 GB disks. Later this year, they are showing off mobile quad core laptops. The limiting factor on boot time and data movement is always the darn HDD, spining at a fixed 7200rpm. Using "parallel-channel" flash SSDs will indeed improve performance significantly, and when I can get my hands on one, I''d be happy to show you numbers and price data. I''ve been installing and testing OS''s on various SSDs and CF devices that are single channel (300x speed equiv for CF marketing), and I can''t wait to test the new parallel channel devices. But in so far as zfs server storage array with heavy write operations? Yeah, we''d have to talk "write data volumes over time" vs. device life span. But that is also set to change, as the vendors are working on newer flash tech that can last much longer. I still see many applications where an SSD or Flash can improve storage system performance in the enterprise. Just stay tuned. Products/solutions are in progress.....> > What about the dual-ported SAS models for enterprise use? > > Bob > =====================================> Bob Friesenhahn > bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ > GraphicsMagick Maintainer, http://www.GraphicsMagick.org/ > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
> Where''s the beef? > > I sense a lot of smoke and mirrors here, similar to Intel''s recent CPU > "announcements" which don''t even reveal the number of cores. No > prices and funny numbers that the writers of > technical articles can''t seem to get straight. > > Obviously these are a significant improvement for laptop drives but > how many laptop users have a need for 11,000 IOPs and 170MB/s?Err, laptop drives? Who cares about laptop drives. 170MB/s writes and 11,000 IOPS? I''ll take four please for my ZFS log device! Seriously, I don''t even care about the cost. Even with the smallest capacity, four of those gives me 128GB of write cache supporting 680MB/s and 40k IOPS. Show me a hardware raid controller that can even come close to that. Four of those will strain even 10GB/s Infiniband. Plus, if Intel are comparing these with 5400rpm drives, and planning them for the laptop market I can''t see them being too expensive. Certainly worth the money. for ZFS. Personally I''d like to see something that mounts on a PCIe card, but if I need to I''ll happily start bolting 2.5" SSD''s to the sides of my server cases! Ross This message posted from opensolaris.org
myxiplx at hotmail.com said:> Seriously, I don''t even care about the cost. Even with the smallest > capacity, four of those gives me 128GB of write cache supporting 680MB/s and > 40k IOPS. Show me a hardware raid controller that can even come close to > that. Four of those will strain even 10GB/s Infiniband.I had my sights set lower. Our Thumper has four hot-spare drives right now. I''d take one or two of those out and replace them with one or two 80GB SSD''s, upgrade to S10U6 when available, and set them up as a separate log device. Now I''ve gotten rid of the horrible NFS latencies that come from NFS-vs-ZIL interaction. It would only take a tiny SSD for an NFS ZIL, really. We have an old array with 1GB cache, and telling that to ignore cache-flush requests from ZFS made a huge difference in NFS latency. Regards, Marion
On Wed, Aug 20, 2008 at 5:29 PM, Tim <tim at tcsac.net> wrote:> > > On Wed, Aug 20, 2008 at 5:17 PM, Bob Friesenhahn > <bfriesen at simple.dallas.tx.us> wrote: >> >> On Wed, 20 Aug 2008, Al Hopper wrote: >> >> > It looks like Intel has a huge hit (product) on its hands with the >> > latest SSD product announcements. No pricing yet ... but the specs >> > will push computer system IO bandwidth performance to numbers only >> > possible today with extremely expensive RAM based disk subsystems. >> > >> > SSDs + ZFS - a marriage made in (computer) heaven! >> >> Where''s the beef? >> >> I sense a lot of smoke and mirrors here, similar to Intel''s recent CPU >> "announcements" which don''t even reveal the number of cores. No >> prices and funny numbers that the writers of technical articles can''t >> seem to get straight. >> >> Obviously these are a significant improvement for laptop drives but >> how many laptop users have a need for 11,000 IOPs and 170MB/s? It >> seems to me that most laptops suffer from insufficent RAM and >> low-power components which don''t deliver much performance. The CPUs >> which come in laptops are not going to be able to process 170MB/s. >> >> What about the dual-ported SAS models for enterprise use? >> >> Bob >> =====================================>> Bob Friesenhahn >> bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ >> GraphicsMagick Maintainer, http://www.GraphicsMagick.org/ >> >> _______________________________________________ >> zfs-discuss mailing list >> zfs-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss > > I don''t know about that. I just went from an SSD back to a SATA drive > because the SSD started failing in less than a month (I''m having troubles > believing this great write-leveling they talk about is working > properly...). And the SATA drive is dog slow in comparison. The biggest > issue is seek times. Opening apps/directories there is a VERY noticeable > difference from the SSD to this drive. > > The user experience is drastically improved with the SSD imho. Of course, > the fact that it started giving me i/o errors after just 3 weeks means it''s > going to be RMA''d and won''t find a home back in my laptop anytime soon. > > This was one of the 64GB OCZ Core drives for reference.Hi Tim. There are a lot of reports of corrupted data with the OCZ Core drive. They have just released a new (replacement perhaps?) product called "Core 2". Its based on a Samsung product. tomshardware.com has a review here: http://www.tomshardware.com/reviews/flash-ssd-hard-drive,2000.html I''m not seeing the promised reliability claims being validated in terms of the warranties that are being offered on most of the current consumer grade SSDs. Regards, -- Al Hopper Logical Approach Inc,Plano,TX al at logical-approach.com Voice: 972.379.2133 Timezone: US CDT OpenSolaris Governing Board (OGB) Member - Apr 2005 to Mar 2007 http://www.opensolaris.org/os/community/ogb/ogb_2005-2007/
On Wed, Aug 20, 2008 at 6:48 PM, Ross <myxiplx at hotmail.com> wrote:>> Where''s the beef? >> >> I sense a lot of smoke and mirrors here, similar to Intel''s recent CPU >> "announcements" which don''t even reveal the number of cores. No >> prices and funny numbers that the writers of >> technical articles can''t seem to get straight. >> >> Obviously these are a significant improvement for laptop drives but >> how many laptop users have a need for 11,000 IOPs and 170MB/s? > > Err, laptop drives? Who cares about laptop drives. 170MB/s writes and 11,000 IOPS? I''ll take four please for my ZFS log device! > > Seriously, I don''t even care about the cost. Even with the smallest capacity, four of those gives me 128GB of write cache supporting 680MB/s and 40k IOPS. Show me a hardware raid controller that can even come close to that. Four of those will strain even 10GB/s Infiniband.How about for serving up CDROM and DVD images (genunix.org). Even two 32Gb drives in a ZFS mirrored config would give you 20K+ read OPs/Sec - as compared to a 10k RPM SCSI drive that starts to fall-over at 400 read IOPS. This type is workload is way over 90% read only - a perfect match for an SSD and this type of workload.> Plus, if Intel are comparing these with 5400rpm drives, and planning them for the laptop market I can''t see them being too expensive. Certainly worth the money. for ZFS. > > Personally I''d like to see something that mounts on a PCIe card, but if I need to I''ll happily start bolting 2.5" SSD''s to the sides of my server cases! >I got to play with one of Sun low-power prototypes and it came with apologies for the way the el-cheapo (transend) SSD was mounted in the case. Wait for it... secured by a two inch wide strip of packing tape. It was certainly a "first" for me! I still smile when I think of it. Apparently the Intel folks put some of their heavy math geeks to work on wear leveling algorithms and came up with something that has advanced the state of the art - from what I''ve heard from people in the know. But few details are available publicly (yet). It''s interesting how the SSD has already turned into a chip based arms raced. Explanation: in the old days, because of the cost and lead time, products were only committed to Integrated Circuits (ICs) aka chips, after the technology had matured. This is no longer the case - as we''ve seen with CMOS digital camera sensors and processing logic. Now we have a case in point with SSDs, where it looks like the technology leaders are out the door with chip based implementations of their first entry into the marketplace. Now thats progress! :) Regards, -- Al Hopper Logical Approach Inc,Plano,TX al at logical-approach.com Voice: 972.379.2133 Timezone: US CDT OpenSolaris Governing Board (OGB) Member - Apr 2005 to Mar 2007 http://www.opensolaris.org/os/community/ogb/ogb_2005-2007/
On Wed, 20 Aug 2008, Al Hopper wrote:> How about for serving up CDROM and DVD images (genunix.org). Even two > 32Gb drives in a ZFS mirrored config would give you 20K+ read OPs/Sec > - as compared to a 10k RPM SCSI drive that starts to fall-over at 400 > read IOPS. This type is workload is way over 90% read only - a > perfect match for an SSD and this type of workload.The logical approach for hot sites like genunix.org is to make sure that the servers are fitted with enough RAM that the disks are virtually idle. I don''t know how many servers are at genunix.org, but if it is just one, then it should definitely have enough RAM to store all the hot data. For example, the BeleniX 0.7.1 ISOs should definitely be in RAM. Perhaps the Genunix Stats page should be updated to list the hardware and software configuration used to serve up all those bytes. Bob =====================================Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/