We once again return to analyze the latest SPECsfs® 2008* benchmark results. There were two new NFS benchmarks, one from EMC on their VNX VG8 gateway with 5-X blades (1 standby) which used SSDs and the other from IBM with their new Scale Out NAS (SONAS) product using disks.  There was also one new CIFS benchmark from EMC using a similar VNX VG8 storage system using more SSD storage.

Latest SPECsfs2008 NFS results

SCISFS110318-001 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

SCISFS110318-001 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 1 SPECsfs2008* Top 10 NFS throughput

Higher is better on this chart. Both new NFS benchmarks broke into the top 10 with blazing throughput performance.

The EMC VNX VG8 with 4-VNX 5700s as its backend, with ~75TB of SAS SSDs and 5TB of 15Krpm disk storage which achieved 497K NFS throughput operations per second.  However, such performance is remarkable regardless of what it took to get there.

Taking a different tack, IBM SONAS used 1975-15Krpm disk drives across 16-storage and 10-interface nodes to achieve 403K NFS throughput operations per second.  Another impressive feat, this time using only lots of disks and servers.  In addition, IBM interface and storage nodes had ~2.4TB of cache memory, of which, 512GB was non-volatile Flash.

SCISFS110318-002 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

SCISFS110318-002 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 2 Top 10 NFSv3 ORT

The EMC VNX VG8 with SSDs also broke into the top 10 in overall response time (ORT) with 0.96 msec..  This is the first NAS system to break the 1 msec ORT barrier for NFS.

Obviously having mostly SSDs on the backend helped here but NFS ORT is also heavily dependent on system overhead. Moreover, ORT is an average of system supplied response times across all NFS workload levels. So an ORT under 1 msec. means that for most of the NFS benchmark response time was under 1 msec. Looking at the chart in the original report[1] EMC’s VG8 VNX system was still supplying under 1 msec. response time at over 295K NFS throughput ops/sec.  One needs to realize that only three other systems are even able to do more than 300K NFS ops per second and none of them appear in the top 10 NFS ORT chart above.

SCISFS110318-003 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

SCISFS110318-003 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

 

 

 

Figure 3 SPECsfs2008* Scatter plot NFS throughput vs. disk drive count

Next we turn to a perennial favorite, our scatter plot showing NFS performance vs. number of disk drives in a system. We exclude all the SSDs from this analysis and to try to show some other components of system performance. One can easily discern the almost 2000 disk drive IBM SONAS system here but the other outlier is the HP BL860c (#3 on the top 10 NFS througput ops/sec chart).

The correlation here is pretty tight, almost an R**2 of 0.9 which says that if you want a high performing NFS system throw lot of disks at it.  According to the linear regression equation, a system with ~2500 disk drives should be able to achieve over 500K NFS ops/sec. beating the EMC VG8 VNX system with 75TB of SSDs.

CIFS analysis

As for CIFS, the new EMC VG8 VNX system used a larger SSD configuration so cannot be compared to their NFS run discussed above. EMC’s VG8 VNX CIFS submission had around 96TB of SSDs with another 5TB of disks.

SCISFS110318-004 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

SCISFS110318-004 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

 

 

Figure 4 SPECsfs2008* CIFS top throughput results

EMC’s VNX VG8 system came in at the new number one with a blistering, over 660K CIFS throughput operations per second.  Almost need a log scale to see the other competitors here.  However, one must realize there are really no other top-end, enterprise CIFS systems submitted yet other than the three EMC systems.  The other two EMC systems had V-Max backends with disk drives and no SSDs. For example, NetApp only shows their midrange systems here.

SCISFS110318-005 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

SCISFS110318-005 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

 

 

Figure 5 Top CIFS ORT results

Once again, we see the EMC VNX VG8 gateway with SSDs as the #1 CIFS ORT. Also when looking at the original report, one can see that the EMC VNX VG8 system sustained over 460K CIFS ops/sec while still responding in under 1 msec. No other CIFS system benchmarked can even achieve that high a throughput level.

Significance

So lots of SSDs can supply amazingly high NFS and CIFS throughput.  The fact that that EMC’s VNX VG8 also was able to achieve great response time is testament to the system effectiveness and other optimizations that allows SSDs to perform so well.

On the other hand if you just want raw throughput, IBM SONAS and HP BL860c have shown that with enough drives and servers one can dial up just about any throughput imaginable.  It’s a pity that few other enterprise class systems have submitted CIFS benchmarks so we can see if a similar strategy works just as well for CIFS systems.

As always we welcome any recommendations for improvement of our SPECsfs2008 analysis.  We also now include a top 30 version of these charts and other charts plus further analysis in our NAS briefing which is available for purchase from our website.

This performance dispatch was originally sent out to our newsletter subscribers in March of 2011.  If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) or subscribe by email and we will send our current issue along with download instructions for this and other reports.  Also, if you need an even more in-depth analysis of NAS system features and performance please take the time to examine our NAS Briefing available for purchase from our website.

A PDF version of this can be found at

SCI 2011 Mar 18 latest SPECsfs2008 benchmark results analysis for NFS and CIFS (PDF 798.3 KiB)

—–

Silverton Consulting, Inc. is a Storage, Strategy & Systems consulting services company, based in the USA offering products and services to the data storage community

 

 

We once again return to analyze the latest Standard Performance Evaluation Corporation Network (System) File System 2008* (SPECsfs® 2008) benchmark results. There were three new NFS benchmarks, two from NetApp (FAS6240 with 1TB Flash Cache and FAS3270) and one from LSI [ONStor] (Cougar 6720). In addition, there were two new CIFS benchmarks one from EMC (Celerra VG8 with VMAX) and one from NetApp (FAS3210 with 512GB Flash Cache).
Latest SPECsfs2008 NFS results

SCISFS101214-001 (c)2011 Silverton Consulting, All Rights Reserved

SCISFS101214-001 (c)2011 Silverton Consulting, All Rights Reserved

Figure 1 SPECsfs2008* NFS throughput vs. memory size
We introduce a new chart for our SPECsfs2008 analysis a throughput vs. memory size scatter plot. For some reason I was very intrigued by this chart when I first created it. Originally all the data was together in one data series with only one linear trend line. I decided to break the data out into two groups, those systems using DRAM caching plus SSDs or NAND caching (6 systems) and those systems using only DRAM caching (33 systems).

After splitting the two types of systems out, I understood better what was shown here. There appears to be a distinct difference in the throughput gained from DRAM caching systems versus SSD use or NAND caching systems. Of course what’s missing from these charts is any comparison of pricing (because it’s not supplied in SPECsfs reports). Also the sample size is very small for the SSD/NAND caching systems and this may skew results.

SCISFS101214-002 (c) 2011 Silverton Consulting, All Rights Reserved

SCISFS101214-002 (c) 2011 Silverton Consulting, All Rights Reserved

Figure 2 SPECsfs2008* NFS throughput operations per second

The other issue with this chart is that SSD use is not accounted for in the memory quantity (just like SPECsfs2008). Adding SSD usable capacity to the memory in a system changed this chart significantly as the SSD system used ~19TB of SSD. Nonetheless, from this chart it seems clear that DRAM caching offers better throughput performance for the same amount of memory. Given today’s limited sample size we cannot discern any statistical difference between NAND cache system versus SSD. This analysis must until more SSD and NAND caching systems report in.

NetApp’s FAS6240 showed up as the new #2 in our top 10 NFS throughput operations per second, the only new submission on this chart. The other thing about the FAS6240 was it use of SAS disks. There haven’t been a lot of NFS results using SAS disk devices so the results here are encouraging.

SCISFS101214-003 (c) 2011 Silverton Consulting, All Rights Reserved

SCISFS101214-003 (c) 2011 Silverton Consulting, All Rights Reserved

Figure 3 SPECsfs2008* NFS top ORT results
Here we show the top 10 NFS ORT (overall response time results). Recall that ORT is a median-like response time over the whole benchmark duration. One can see all of the new NFS benchmarks in this chart. The NetApp 6240 showed up as the #1 with an ORT of 1.17 msec., the FAS 3270 came in at #7 with an ORT of 1.66 msec, and the LSI (ONStor) Cougar 6720 came in at #9 with an ORT of 1.67 msec.

I would have expected the NAND Cached FAS6240 to do well with ORT but the Avere Systems continue to amaze. However I must say that the Avere systems at #2, 3 and 5 may enjoy some advantage due to their relatively large DRAM caches (98, 163, and 424 GB respectively).

CIFS analysis

Below we report on top CIFS throughput results which include the latest submissions from EMC and NetApp.

SCISFS101214-004 (c) 2011 Silverton Consulting, All Rights Reserved

SCISFS101214-004 (c) 2011 Silverton Consulting, All Rights Reserved

Figure 4 SPECsfs2008* CIFS top throughput results

EMC’s Celerra VG8 came in as the new #1 with ~143K CIFS throughput operations per second and the NetApp FAS3210 came in at #3 with ~65K CIFS throughput operations per second. The strange thing about the top three results is that the #2 and #3 results used SSDs and NAND cache respectively but the #1 result just had a lot of disks (312). As we have said in the past, SSDs or NAND cache can substitute and perform just as well as more spindles if you don’t need the capacity.

SCISFS101214-005 (c) 2011 Silverton Consulting, All Rights Reserved

SCISFS101214-005 (c) 2011 Silverton Consulting, All Rights Reserved

Figure 5 Top CIFS ORT results
Here we show the top 10 CIFS ORT (overall response time) results. Similar to NFS ORT results, CIFS ORT is a median-like response time over the whole benchmark duration. Both new submissions showed up in the top 10, NetApp FAS3210 at #4 and EMC Celerra VG8 at #9. Not surprising is the fact that 3 out of the top 4 ORT systems used FlashCache. Apple still holds the coveted #1 spot with an ORT of 1.22msec., but others are starting to encroach.

Significance

I struggled to understand the NFS memory size vs. throughput scatter plot chart. It should have been plain to see but it was unclear until I split out the data into two series. The fact that DRAM provides better throughput than NAND or SSDs is pretty noticeable, but lacking cost information, it’s impossible to compare cost effectiveness.

As always we welcome any recommendations for improvement of our SPECsfs2008 analysis. We also now include a top 30 version of these charts and other charts plus further analysis in our NAS briefing which is available for purchase from our website.

This performance dispatch was originally sent out to our newsletter subscribers in December of 2010.  If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) or subscribe by email and we will send our current issue along with download instructions for this and other reports.  Also, if you need an even more in-depth analysis of NAS system features and performance please take the time to examine our NAS Briefing available for purchase from our website.

A PDF version of this can be found at

SCI 2010 Dec 14 latest SPECsfs2008 benchmark results analysis for NFS and CIFS (PDF 721.6 KiB)

 

Silverton Consulting, Inc. is a Storage, Strategy & Systems consulting services company, based in the USA offering products and services to the data storage community


* All results from www.spec.org as of 14 Dec 2010

 

 

We now turn to analysis of the latest SPECsfs® 2008* benchmark results. Fortunately there were a number of new NFS and CIFS benchmarks over the last quarter, including eight new NFS results in the top 10 and a couple of new CIFS results in the top 10.

Latest SPECsfs2008 NFS results

(SCISFS091230-001) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS091230-001) 2009 (c) Silverton Consulting, All Rights Reserved

Huawei Semantec’s N8500 NAS storage system came in with a new number one result at over 176K NFS throughput operations per second.  This was accomplished on Huawei server and storage hardware running Semantec’s FileStore Clustered NAS appliance software.  Rounding out the rest of the top 3 were BlueArc’s new “midrange” Mercury 100 cluster, and HP BL860C.  Both the HP and Huawei Semantec systems supported multi-node clusters, 12 nodes and 4 nodes respectively and both were running Semantec’s VxFS software.  The BlueArc system has hardware acceleration and running on a two-node cluster.

Avere is a new NAS storage system and also supports a multi-node cluster.  There were three Avere systems benchmarked in this last quarter with six nodes, two nodes, and one node.  Their six-node system attained #4 of the top 10, running at over 130K NFS throughput operations with minimal disk drives (using only

79 drives).  Most of the other systems in the top 10 had many more drives with the exception of NetApp FAS3160 with PAM acceleration that ran with 56 drives.

(SCISFS091230-002) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS091230-002) 2009 (c) Silverton Consulting, All Rights Reserved

For operational response time (ORT) results, we can see more of Avere systems in the top results attaining three out of the top four ORT results at 1.3, 1.33, and 1.38 msec. respectively.  It’s interesting to see that the other top 10 NFS throughput results from Huawei Semantec and the NetApp systems also achieved top 10 ORT results as well.

Latest SPECsfs2008 CIFS results

(SCISFS091230-003) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS091230-003) 2009 (c) Silverton Consulting, All Rights Reserved

There were two new CIFS benchmarks from Fujitsu submitted this last quarter, one for their Primergy TX200 S5 and one for their Primergy BX920 S1.  Both results managed to attain a position in a top 10 CIFS result, one in throughput and the other in ORT.

It doesn’t appear that the SPECsfs CIFS benchmark is gaining much mindshare.  As of this dispatch there are only 11 total submissions.  Nonetheless, the latest Fujitsu Primergy TX200 reached the top 10 (out of 11 – sigh).   The dominant result remains with Apple’s Xserve running Snow Leopard running over 44K CIFS throughput operations per second.

(SCISFS091230-004) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS091230-004) 2009 (c) Silverton Consulting, All Rights Reserved

In the CIFS operational response time results, the new Fujitsu Primergy BX920 S1 attained #5 with a 2.9 msec. ORT.  Once again an older Apple storage system holds the top ORT result at 1.2 msec.

Significance

Nice to see some activity at the top end in NFS results.  The Huawei Semantec system has laid down the gauntlet with some pretty impressive numbers but we have yet to hear from some other vendors with enterprise class systems.  NetApp, as always, was early to submit benchmark results for their current systems and their next generation Data ONTAP 7G will also support a clustered file system. Can’t wait to see how well that performs.

We are a bit disappointed in the paucity of CIFS results and yet continue to report them in the hope that more will be released.  But as the benchmark has been out for over 18 months now it is not gaining many adherents.  Also it would be wonderful to see more submissions for both benchmarks using the same hardware/software so that end-users could see what they are getting when they use CIFS or NFS.

This performance dispatch was originally sent out to our newsletter subscribers in December of 2009.  If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) or subscribe by email and we will send our current issue along with download instructions for this and other reports.  Also, if you need an even more in-depth analysis of NAS system features and performance please take the time to examine our NAS Briefing available for purchase from our website.

A PDF version of this can be found at

SCI 2009 December 31 Update to SPECsfs® 2008 performance results (PDF 578.6 KiB)

Silverton Consulting, Inc. is a Storage, Strategy & Systems consulting services company, based in the USA offering products and services to the data storage community


* SPECsfs2008 results from http://www.spec.org/sfs2008/results/

 

We now turn to analysis of the new SPECsfs®2008* benchmark results. Unfortunately there were not a lot of highend SPECsfs2008 results, most notably ONStor and Apple for NFS, and Apple and Fujitsu Siemens for CIFS.

Latest SPECsfs2008 results

(SCISFS090625-001) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS090625-001) 2009 (c) Silverton Consulting, All Rights Reserved

ONStor Cougar now takes second spot in the top 10 throughput results.  The Cougar system had ~½ the disks of the ExaStore box and ~7 times less memory (cache).  Given all that, its results standup pretty well.  The two new Apple NFS benchmark results (Snow Leopard and Leopard server) round out the rest of the new members to the top 10 list at numbers 7 and 10 respectively.

Recall from our last report# that some NetApp results utilized their PAM card. Also, the SGI product result used Infiniband, both ExaStore benchmarks used10GbE and all the rest  use GigE.  In all fairness the networking connection may not be a limiting factor in SPECsfs2008 results.

(SCISFS090625-002) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS090625-002) 2009 (c) Silverton Consulting, All Rights Reserved

As discussed last time for NFS ORT results, one can clearly see the advantage of NetApp’s PAM with FC disks and yet, the new ONStor Cougar benchmark shows up at number 3, only ~60 microsec behind the NetApp/PAM result.  The only other new showing was Apple’s Snow Leopard server coming in at number 9.

Next we turn to CIFS results, the five new results have more than doubled SPECsfs2008 CIFS benchmarks.   Recall the SGI is using Infiniband while all the others use GigE interfaces.

(SCISFS090625-003) (c) 2009 Silverton Consulting, All Rights Reserved

(SCISFS090625-003) (c) 2009 Silverton Consulting, All Rights Reserved

We suppose it’s not surprising to see Apple’s Snow Leopard leading the pack, coming in at the new #1 in CIFS throughput considering its market place but one would think others could do better.  More impressive is that the Snow Leopard result used only 65 disks whereas the SGI result sported 242 disks (~4X).  It’s unclear to us whether this is the new Apple OEM of Sun ZFS file system at work here, but clearly Apple CIFS performance has improved significantly.

(SCISFS090625-004) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS090625-004) 2009 (c) Silverton Consulting, All Rights Reserved

Once again, Apple shows up well in CIFS ORT results.  Although, as best we can determine this #1 result was an early Leopard version (Mac OSX10.5.1) whereas the #3 result (using Mac OSX10.5.7) had a 2.93Ghz Nehalem processor.  The other major difference was a dual port GigE card for the #1 result vs. a 6-port GigE card in the slower version.

(SCISFS090625-005) 2009 (c) Silverton Consulting, All Rights Reserved

(SCISFS090625-005) 2009 (c) Silverton Consulting, All Rights Reserved

Figure 5 SPECsfs2008* CIFS vs. NFS throughput correlation

We have discussed this in earlier reports but once again the results would support our contention that the CIFS protocol results in better throughput than NFSv3.  As pointed out to me, a couple of provisos are warranted here, namely:

  • NFS workloads are not readily comparable to CIFS in a number of dimensions not the least of which is that NFS is stateless and CIFS is state-full.   Also, the relative proportions of the actual workloads don’t exactly matchup, e.g. percentages for NFS read and write operations versus CIFS read_andx and write_andx operations are slightly different (NFS read@18% vs. CIFS read_andx@20.5% and NFS write@10% vs. CIFS write_andx@8.6%), file sizes are different, and all the remaining operations, which, to be fair, represent a significant majority of their respective workloads, are by definition, nigh impossible to compare. SPECsfs benchmarks for the two are implemented to reflect all of these differences.
  • A majority of these results (3 of 5) come from the same vendor (Apple) and their great CIFS and/or poor NFS implementations may be skewing results.
  • Only five subsystems have recorded results for both interfaces but the correlation looks pretty good for now.
  • Normally, host operating system affects could skew these results but the SPECsfs2008 benchmarks emulate their own client side stacks for both protocols, thus negating any operating system affects.

Nonetheless, once again, considering that at the user level all specific protocol details result in emulating comparable end-user workloads, the results do show a significant advantage for CIFS (~2.4X) throughput over NFS.

Significance

Our earlier discussion on CIFS vs. NFSv3 throughput differences resulted in quite a lot of discussion.  It was early then, and still is now, but we continue to stand by our claim, given benchmark results, CIFS seems to perform better than NFSv3.  More dual protocol results should help clarify this relationship.

Slowly, more SPECsfs2008 results are being released.  But, where are the major NAS systems.  It’s been 10 months since the old SPECsfs benchmark was retired and we still lack benchmark results for all the major NAS systems.  In the mean time, smaller players continue to release results; just happy to get any visibility, validity and traction they can muster.

This performance dispatch was originally sent out to our newsletter subscribers in June of 2009.  If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) or subscribe by email and we will send our current issue along with download instructions for this and other reports.  Also, if you need an even more in-depth analysis of NAS system features and performance please take the time to examine our NAS Briefing available for purchase from our website.

A PDF version of this can be found at

SCI 2009 June25 Update to SPECsfs® 2008 performance results (PDF 675.4 KiB)

Silverton Consulting, Inc. is a Storage, Strategy & Systems consulting services company, based in the USA offering products and services to the data storage community

* SPECsfs2008 results from http://www.spec.org/sfs2008/results/

# Available at http://www.silvertonconsulting.com/page2/page2d/storage_int_dispatch.html

For a no risk, free 30 minute consultation please give us a call at +1-720-221-7270

If you are interested in receiving our free monthly newsletter on major storage announcements and storage performance results please subscribe by email
Silverton Consulting, Inc., P.O. Box 492
Broomfield, CO, 80038-0492, USA
© 2004-2012 Silverton Consulting, Inc., All Rights Reserved
Contact|SCI Privacy Policy|Sitemap
Suffusion theme by Sayontan Sinha