We return now to our analysis of the latest SPECsfs® 2008* benchmark results[1]. There have been six new NFS and no new CIFS/SMB submissions since our June report.  The new NFS submissions included the following: an EMC VNX8000 with 8 X-Blades (1 standby), three Hitachi systems, the Hitachi NAS (HNAS) 4100, dual node gateway, the Hitachi Unified Storage (HUS) 4060 and HUS 4080 and two Oracle Systems, the Oracle ZFS Storage ZS3-2 and ZFS Storage ZS3-4. The new submissions changed a number of our top ten charts, which are reviewed below. Most of our other SPECsfs2008 charts are viewable in previous dispatches available from our website.[2]

Latest NFS results

Image may be NSFW.
Clik here to view.Figure 1 SPECsfs2008 NFSv3 Top 10 NFS Overall Response Time (ORT)

Lower is better in Figure 1. There were three new systems in our Top 10 NFS ORT chart. The Oracle ZFS ZS3-4 is our new number one at 0.70msec. the EMC VNX8000 came in at third place with 0.78msec. and the Oracle ZFS ZS3-2 came in at eight place at 1.12msec. All the new systems in Figure 1 had some flash storage, the two Oracle systems used 8-73GB SSDs with lots of disk and the EMC VNX8000 was an all-flash array with 544-200GB flash drives. The ZFS Storage used their flash for ZFS Intent

Log (ZIL). As I understand it, the ZIL is used to buffer write activity (<64KB of data) and as a log for all file management operations. How Oracle ZFS storage is able to respond so well with just a relatively small amount of flash is beyond me.

There’s only one all disk system left in our Top 10 NFS ORT chart and that being the Huawei Symantec Oceanspace N8500 Clustered NAS, which came in at seventh place and used 960 disk drives with no flash but with over a TB of DRAM memory in the cluster. The fact that seven out of the top ten systems above had an ORT of under 1msec. is amazing, but is a significant proof point that flash can deliver faster response times.

Next we turn to NFS throughput operations per second.

Figure 2 SPECsfs2008 Top 10 NFSv3 throughput ops/sec.Image may be NSFW.
Clik here to view.

Higher is better in throughput results. The only new system here is the EMC VNX8000 all-flash array, which came in at tenth place with just under 600K NFS throughput operations per second. Hybrid SSD-disk and all-flash systems seem to dominate here with the lone exception being once again the Huawei Symantec N8500 NAS cluster.

We now turn to normalized NFS operations per second. First, we examine NFS operations per second normalized by disk spindle count.

Figure 3 SPECsfs2008 Top 10 NFS throughput operations per second per disk

Image may be NSFW.
Clik here to view.

Looking at just disk only submissions we find another new submission the HNAS Model 4100 gateway, which had a Hitachi VSP behind it.  The HNAS-VSP combination used 480-146GB disk drives but generated over 290K NFS throughput ops/sec., which placed them fifth on the above chart with slightly over 600 NFS ops/sec/disk spindle.  The two new HUS systems didn’t show up quite as well but came in the Top 20 version of this chart which is only available in our SAN-NAS Buying Guide purchasable on our website.

Next we normalize NFS operations by the amount of flash in use.

Figure 4 SPECsfs2008 Top 10 NFS throughput operations per second per GB of flashImage may be NSFW.
Clik here to view.

I have tried to normalize hybrid and all-flash activity by both SSD/flash card and amount of flash but nothing comes close to the correlation we see with disk drives in disk only storage. Nonetheless, it’s been a long time since we showed this chart and two new systems showed up in the top ten here, the new Oracle systems coming at second and third place with ~770 and ~360 NFS operations per second per GB of flash. Again, they do very well with a relatively modest amount of flash. NetApp is the other solution that does well with small amounts of flash, ranging in the systems above from 1TB to 12TB (24-node) of flash. The Avere FXT 3500 is another matter, they had a small amount of flash, 800GB, but used over 6TB of DRAM memory as well, which is even faster than flash.

Significance

Still no new CIFS/SMB results, I keep wondering if maybe SPECsfs (2014?) is moving to SMB2 or even SMB3, which may be why everyone is holding off but I have no official confirmation on this.

The six new NFS results for this quarter seemed to do well in most of our charts. Oracle’s new ZFS storage systems did surprisingly well on ORT, HNAS-VSP combination placed well in NFS ops./sec./disk spindle and the brand-new, all-flash VNX8000 did well both on ORT and NFS ops./sec.

We provide more detailed file system performance and NAS system feature analysis plus our SPECsfs2008 ChampionsCharts™ for NFS and CIFS in our recently updated (September, 2014) NAS Buying Guide available from our website . Also Top 20 versions of some of these charts are displayed in our recently updated (October, 2014) SAN-NAS Buying Guide also purchasable from on our website.

We are always interested in improving our performance analyses for SPECsfs2008 as well as other performance benchmark results. If you have any ideas as to how to enhance this analysis or ways to supply better insight into file or storage system performance, please don’t hesitate to drop us a line (contact information is at bottom left).

~~~~

[This performance dispatch was originally sent out to our newsletter subscribers in September of 2013.  If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) and we will send our current issue along with download instructions for this and other reports. Dispatches are posted to our website at least a quarter or more after they are sent to our subscribers, so if you are interested in current results please consider signing up for our newsletter.]  

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