SPC-1*results

We start our discussion with the top 10 IOPS performance metric for SPC-1.

(SCISPC111122-001) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 1 Top 10 SPC-1* IO/Sec

Higher is better on the IO/Sec or IOPS™ results.  Here we can see the new HP 3PAR V800 storage with 8-nodes coming in at #1 and the HDS VSP with 8-virtual storage processors showing up as #10.  One thing not evident in this comparison is that the 3PAR system used 1920 drives while the HDS VSP only had 1152. This difference would normally show up better in IOPS/spindle but as none of the new system results broke into that metric’s top 10, it is not shown in this report.

Another thing not readily apparent was that the HDS VSP system LUNs were thinly (dynamically) provisioned.  I believe this is the first SPC-1 report using thinly provisioned volumes and expect that thin provisioning may degrade performance.

Next let’s turn to IOPS/$/GB.

(SCISPC111122-002) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 2 IOPS/$/GB

Higher is better on IOPS/$/GB.  Here we can see all three systems the #1 HP 3PAR V800 storage, the #7 Oracle Sun ZFS Storage 7420c Appliance and the #10 HDS VSP.  I prefer this approach to measuring the cost of storage IO rather than SPC-1’s reported measure of IOPS/$™ as it incorporates the size of backend storage.

One other thing not discernable in the above is that the Oracle Sun ZFS Storage 7420c used SSDs and had a combination of 8-73GB SSDs write flash accelerators, 8-512GB SSD read flash accelerators and 280-300GB SAS drives.  Most of the other systems here only used disk drives, except for the TMS RamSan-630 that used all SSDs.

(SCISPC111122-003) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 3 SPC-1 IOPS vs. drive spindles scatter plot

 Next we have a favorite SPC-1 char that shows a scatter plot of IOPS vs. LRT with bubble size equal to system cost ($K).  Here one can see all three new system results, i.e., the HP 3PAR V800 system at the top of the chart, the HDS VSP somewhere in the middle of the pack around 270K IOPS, and the Oracle Sun ZFS just below the 150K IOPS line.

(SCISPC111129-004) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 4 Scatter plot: IOPS vs. number of disk drives

Finally, we include our occasional IOPS vs. drive count scatter plot.  We like to show this to help determine whether subsystem IOPS performance is driven purely by disk count.  Alas, as shown above, a R**2 of 0.96 clearly indicates that SPC-1 IOPS rate is driven by spindle count.

A couple of caveats for this chart, we threw out any result with SSDs or multiple drive capacities so as not to skew the results. Also this only contains results for drives larger than 140GB (SPC-1 results go all the way back to 36GB disks).

I like to think that the wider variance at higher IOPS rates is telling me something.  Consider the column of 3 results centered about 2000 disk drives.  The lowest performing system was a 4-node IBM SVC 5.1 system, the next higher performer was a 6-node IBM SVC 5.1 configuration and the top system was the 8-node HP 3PAR V800 storage system.  It now seems obvious here but at some constant level of disk drives, the more processing power you throw at SPC-1 the more IOPS you can attain.

This may explain the relatively narrow grouping at fewer than 500 drives. Most of these results probably just used a dual controller configuration for their storage systems.

Significance

It’s good to see some new SPC-1 results especially at the high-end, enterprise class level.  With these systems the IOPS rate just continues to climb and there doesn’t appear to be any real physical limit to storage performance.

In addition, at least when it comes to IOPS disk drive only systems seem to be holding their own against hybrid SSD-disk systems and all SSD systems.  However, IO latency or LRT™ (not shown in today’s dispatch) or is where all SSD systems shine, holding the top 5 spots and have yet to be usurped by anything than other SSD systems.

As always, suggestions on how to improve any of our performance analyses are welcomed.  Additionally, if you are interested in more details, we now provide a fuller version of all these charts in SCI’s SAN Storage Briefing which can be purchased separately from our website[1].

[This performance dispatch was originally sent out to our newsletter subscribers in November 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 SAN storage system features and performance please take the time to examine our SAN Storage Buying Guide available for purchase from our website.]

~~~~

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

FAS2240 system

The FAS2240 is a new, entry-level storage system.  The system lists at under US $16K and has an impressive list of software functionality, including just about any advanced storage system feature imaginable.  This new system also introduces some interesting new channel capabilities such as storage capacity additions and software licensing at PoS.

In addition, NetApp is retiring their FAS2020 at this time.  The FAS2020 had been the low-end of their storage line, which now makes the FAS2040 their new low-end storage system.

FAS2240 Hardware

The new system supports up to 374TB  (-2) or 432 TB (-4) in a single or dual controller configuration.  Each system has up to 12GB of cache memory for dual controller configurations.

There are two different form factors for the FAS2240 a 2U utilizing SAS high performance disks and a 4U featuring high capacity SATA disks.  The new product is focused on channel delivery and drive shelves can be added in the channel at point of sale and add-on software functionality can easily be license enabled as well.

In addition, the FAS2240 disk storage can be converted to disk shelves for higher end NAS heads such as the FAS3210 when the time comes to convert to higher end storage.  This provides an in-place data migration, where the aggregate data doesn’t actually move but ownership of the aggregate changes from the FAS2240 to the FAS3210 controller head.

The FAS2240-2 comes with 12 or 24 450 or 600GB SAS disks and the -4 comes with 12 or 24 1, 2, or 3TB SATA drives.  In addition to the internal disk drives, external drive shelves can be attached to increase capacity.  The maximum number of disk drives for both systems is currently 144.

Both systems can be ordered with a single or dual controller option.  Also, the base system comes with 4 or 8-GigE interfaces, for single and dual controller systems respectively, and optional I/O cards with up to 4-8GFC or 4-10GbE interfaces.

FAS2240 Software

The FAS2240 only supports Data ONTAP 8.1 or beyond and comes bundled with all Data ONTAP Essentials including: storage efficiency features (FlexVol®, data deduplication, data compression, and thin provisioning), availability features (Multi-path IO, SyncMirror, MultiStore®), data protection (RAID-DP, Snapshot, and Open Systems SnapVault®), performance (FlexShare ®) and management (System Manager, Ops Manager, Protection Manager, and Provisioning Manager), all at no additional charge. In addition, all storage protocols are bundled into the base price of the system, although if you want FC you have to purchase an optional IO card.

If customers want even more functionality they can purchase up to 7 options of software including

• SnapRestore® – faster restore of snapshot copies,
• SnapMirror® – Synchronous, semi-synchronous, and asynchronous data replication for flexible disaster recovery
• SnapVault® – disk to disk backup software for copying data to secondary storage systems,
• FlexClone® – instant virtual copies of databases or VMs,
• SnapManager Suite® – application and virtual machine backup, recovery and cloning available for Oracle®, Exchange, SharePoint®, SQL® server, SAP®, Virtual Infrastructure and Hyper-V™
• OnCommand Insight Balance for NetApp – advanced analytics for physical and virtual environments
• Complete Bundle – includes all of the above, except Insight Balance for NetApp.

Announcement Significance

NetApp rolled out their “Smart Decisions” theme with this announcement.  Storage Smart Decisions start right with unified storage at under US $7,500K (FAS2040), keep it simple with powerful management tools, and grow- smart with an architecture that protects customer investment.  The FAS2240 will become their primary vehicle to take storage Smart Decisions out to the world.

I suppose it was time for NetApp to refresh some of their low-end storage. With the introduction of the FAS2240, NetApp will retire the FAS2020 and some FAS2040 legacy configurations.  Recall that previously, NetApp bundled more of their software solutions on the FAS2040 and the FAS2240 take this another step further.

Obviously, the FAS2240 is primarily destined for the channel. To that end, all that software bundling, ease of PoS storage capacity will make this an even more channel friendly product.

[This storage announcement dispatch was originally sent out to our newsletter subscribers in November 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.]

~~~~

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

This Storage Intelligence (StorInt™) dispatch covers recent Microsoft Exchange 2010 Solution Review Program (ESRP)[1] v3.0 results. Since we last discussed this 1K-and-under mailbox category nine months ago, there have been seven new ESRP submissions. Future dispatches will report on the 1K-to-5K and over-5K mailbox categories, but all previous ESRP V2 and ESRP V3 performance dispatches are available on SCI’s website[2].

Latest ESRP V3.0 results

We start our ESRP analysis with Exchange database and log access latency results.  Recall that this chart is sorted by database read latencies.

(SCIESRP111029-001) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 1 Top ESRP database-read latencies

For Figure 1, the 3-HP systems (#8-10) and the 2-EMC VNXe systems (#3 & 6) are new here. ESRP database and log latencies remain a favorite ESRP metric because it correlates so well with end-user experience.  However, I coming to the belief that latency is driven by disk speed.  For example, the first three systems above had 15Krpm drives, and all the rest were 7200 or below except for the other EMC VNXe 3100 (#6) that was running 15Krpm drives.  Unclear why it had such poor read latency, but maybe it was overtaxed.

We now turn to ESRP database transfers per second.

(SCIESRP111029-002) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 2 Top 10 Database transfers per second

Similar to the above all the HP and EMC VNXe runs were new to this chart. As discussed previously the Xiotech system seems to dominate this metric but at least we can see one system that is almost 1/2 as good (#2 HP P2000 G3).   Here we have a nice mixture of FC (#1), iSCSI (#2,6,7,8&10), and SAS attached (3,4,5&9) storage systems.

Next we examine Exchange log playback performance.

(SCIESRP111029-003) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 3 Top log playback times

This is one of my favorite ESRP metrics because:

• It’s independent of Exchange and ESRP configuration parameters, i.e., it uses similar 1MB log files for any and all storage subsystems
• It’s a mixture of log and database processing that’s done as fast as the storage system and server(s) can accomplish it.

There are seven new submissions on this chart, i.e., everything except for the EMC Celerra (#5), Xiotech (#6) and EMC CLARiiON (#8) is new.  The NetApp FAS2040 had a surprisingly good showing here just edging out the HP P2000 G3 iSCSI system.

For no apparent reason, ISCSI systems (#1-5, & seem to dominate log playback, the Xiotech is FC and the remainder are SAS attached systems. Also drive speed doesn’t seem to have any impact here as the top 2 and bottom 4 use 7200rpm drives, while the middle 4 use 15Krpm (#3,4, & 6) or 10Krpm (#5) drives.

Unclear why the two top systems were able to process logs so fast with such slow disks. But it may have something to do with the sequential nature of log processing which could negate the drive speed advantage.  There’s more to this metric than is obvious to the casual observer.

We have discussed this before but another of my favorite ESRP reported metrics is DB backup activity. ESRP publishes two measurements on database backup activity 1) Data backed up as an average MB/sec/database and 2) Data backed up as an average MB/sec/server.  Neither backup metric is entirely sufficient from our perspective so we have computed a new one.  Specifically, we multiply the average MB/sec/database times the number of databases to calculate a total database backup in MB/sec.

(SCIESRP111029-004) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 4 Top 10 total database backup rates

Figure 4, summarizes our new total database backup rate metric and ranks the top 10 storage systems in this category.  One would think that more disk drives would provide better backup performance and for this category this seems to be the case.  The #1 system had 78 disk drives used in a RAID10 configuration which seemed to help it beat the competition.  The #2 (Xiotech) system had 20 active disk drives and the remainder of these systems all 12 or fewer drives.

Finally, we discuss the database transfers per spindle metric.

(SCIESRP111029-005) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 5 Total DB transfers/second/disk drive

Once again Xiotech dominates this category but the other surprise is the very same VNXe 3100 with 1Kmbx we thought was overtaxed in the read latency chart (see Figure 1 above).  There seems to be nothing wrong with its showing here, but this may be due to the fact that it was only running 7 drives.

Conclusions

Recently, we were asked what 3 metrics to use for measuring ESRP subsystem performance. Of course we responded with 4: read latency, log playback, total database backup and database transfers per spindle.  Note that only two of these are actually reported in ESRP reports.  However, we would probably now add normalized database transfers per second.

Absolute (non-normalized) numbers were used in this 1K-and-under mailbox results analysis but we would typically be reporting on database transfers per 1000 or 5000 mailboxes just to remove the correlation to mailbox count.  Nevertheless, we reserve the right to use whatever metric we can to analyze future ESRP runs.

Finally, as discussed in prior dispatches, ESRP/Jetstress results seem destined to be difficult to compare but in our view, merit the effort.  As such, feel free to contact us with any constructive ideas on how to improve.  In that regard, our contact information can be found below or on our website at SilvertonConsulting.com.

[This performance dispatch was originally sent out to our newsletter subscribers in October 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 SAN storage system features and performance please take the time to examine our SAN Storage Buying Guide available for purchase from our website.]

~~~~

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

ACE federated file systems

With the new ACE capabilities file data can now be shared across a number of SONAS clusters located around the world, creating a federation of SONAS systems.  By doing this, data located at any of the federated SONAS sites can now be readily accessed by any other location in the federation.

It starts with stitching together a global file system, spanning multiple SONAS clusters. With the global file system across multiple SONAS clusters, access can then be provided to all the files underneath this system without having to change OS support.

In this way, with a federation of SONAS clusters located in Hong Kong, London, and New York, personnel at each site could have access to data located in other sites.  However, a single site will be designated to have write ownership over a fileset (i.e. directory/folder) level, although such ownership can be transferred. Many filesets can be deployed for granularity.  In addition, sites that have read access can copy files into a different fileset or locally to a users’ system. When data is being written at a cache site, it automatically will transfer any changes to a home site after which the other sites would once again have access to the most current version of the data.

In addition to providing global file access, ACE can be used to preposition files to any federated cluster location as needed.  This can be especially useful for more predictable access to files from other locations.  When it’s known that a certain set of files will be needed at an alternate site, ACE policies can be established that will pre-position that data at the other site such that access will appear to be the same speed as local file access.

All this differs substantially from storage or file replication services. With file replication, there can be only one reader/updater present in the replication cluster. In contrast, with ACE federation any cluster location can access data wherever it’s needed.  Moreover, ACE policy management can also restrict certain directories to be only accessible, locally if so desired, blocking them from the global file system. In this way, subsets of cluster data could be available globally and other data could be restricted to local only access.

ACE policy management services

In addition to creating a SONAS federation, ACE can also provide other automated policy management for the files it controls.  Specifically, ACE policies can be created that identify which files have been modified since last backup, need to be replicated, or have expired and need to be deleted.  Together with its automated policy management ACE can be used to more effectively manage, protect and replicate file data.

Storwize V7000 Unified

In addition to the ACE announcement described above, IBM also released a new packaging together of Storwize V7000 and SONAS into a single, bundled offering.  In this case Storwize V7000 and two SONAS interface nodes are packaged together and use the same storage for both file and virtualized block storage services.

However, currently the ACE capabilities of Storwize V7000 unified are restricted to a single site but one can readily see how ACE federated file systems could be extended to use V7000 unified clusters just as well.  Nonetheless, ACE capabilities for a single site, described above, can provide some unique capabilities in and of itself.

The Storwize V7000 Unified file (SONAS interface) modules provide 2-10Gbps Ethernet file interfaces and consume 2-8Gbps FC interfaces to the Storwize V7000 control enclosures.  The Storwize V7000 Unified still supports up to 8-8Gbps FC host ports, 4-GigE interfaces and optionally, 4-10Gbps iSCSI port interfaces.

Announcement Significance

I am aware of large design shops that have for years used FTP to move data from one design lab to another around the world, to keep their design work” following the sun”.  These processes involve elaborate scripts and full directory migration from one lab to another, taking place off hours.  With the introduction of ACE federation, this sort of capability can be completely automated through the use of policies to preposition the data at each lab over time. However, this wouldn’t even be needed as designers at the sites around the world could just access the files as needed.  Of course there would be more overhead to access the data from other SONAS locations but this could be a tradeoff -  better performance by pre-positioning file data or less bandwidth use with dynamic access driven movement. In any event, with SONAS ACE federation, at least you will have a choice.

The Storwize Unified follows a trend in the industry to provide the option for both file and block storage from a single storage system, especially appealing to midrange customers.  While Storwize V7000 Unified ACE capabilities may be limited to non-federated use today, that restriction will be removed over time.

There seems to be a storage industry move afoot to stitch together multiple data centers into one world wide, universal data center.  While private cloud capabilities make this easier to do for applications and servers, one vexing problem has remained namely, how to make data available across these multi-site environments.  EMC VPLEX and IBM’s ACE are responses to these issues and we are certain more will come.

[This storage announcement dispatch was originally sent out to our newsletter subscribers in October 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.]

~~~~

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

Latest SPECsfs2008 results

(SCISFS110929-001) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 1 Scatter plot of “NFS throughput” per disk vs. “CIFS throughput” per disk

We made a mistake on our previous version of this chart and as such, have fixed and updated this chart since last time.  Mostly the difference between this version and the original is the removal of multiple EMC Isilon NFS and CIFS runs and of course the addition of the latest HDS NAS 3090-G2 NFS submission somewhere in the lower left corner.

As it turns out Isilon had been using SSDs all along in their submissions that we didn’t catch before.  Don’t know how we missed this in Isilon’s benchmark reports as they clearly indicated the use of SSDs. Be that as it may, sorry for any confusion we may have caused.

The correlations between number of disks and protocol throughput are still pretty good with 0.98 for CIFS and 0.82 for NFS.  Although without the EMC Isilon CIFS submissions we really only have 15 CIFS vs. 37 NFS submissions and the CIFS results are mainly skewed to low-end systems.  Nonetheless, the results are still pretty impressive and clearly show an advantage for CIFS, at least with respect to throughput per disk spindle deployed

(SCISFS110929-002) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 2 Top 10 NFS throughput operations per second per disk drive

Higher is better on this chart. One can see the newest entry here, the HDS 3090-G2 BlueArc system coming in at #10.  BlueArc (and it’s former HDS OEM, now parent company) have 7 of the top slots here and Avere has the rest.  The results above seem to indicate that the HDS 3090 system is using BlueArc’s Mercury or midrange controller but looking at the reports it could just as easily have been the Titan or high-end controller.

As you may recall Avere system is a NAS virtualization engine which has other NAS boxes behind it.  One thing about this chart is that we exclude any and all SSDs or NAND based caching. But in all honesty, the Avere systems have lot’s of cache (163GB & 424GB of RAM for #1 & 2 respectively) and the latest HDS entry is no slouch either with 184GB of RAM caching spread throughout the VSP and BlueArc controllers.  We may need to establish a cutoff limit for RAM as well here.

Even though there have been no new CIFS submissions, we provide a similar chart below to Figure 2 because we have not shown one recently.

(SCISFS110929-003) (c) 2011 Silverton Consulting, All Rights Reserved

Figure 3 Top 10 CIFS throughput operations per second per disk drive

In Figure 3 the Fujitsu TX300 S5 RAID 50, Apple’s Xserve, and Fujitsu TX300 S5 RAID0 (ouch!) take top honors here with 20, 65 and 20 disks respectively.  In contrast, EMCs Celerra VG8 had 280 disks and the Huawei Symantec system 1344, considerably more drives than the other entry-level systems elsewhere on this chart.

Significance

We believe CIFS is still winning in its horse race against NFS but future submissions could tip the scales either way.  More CIFS submissions, especially at the enterprise level (and without the use of SSDs) would help.

As an aside, we went looking to determine what version of SMB (CIFS) SPECsfs2008 used but there was none stated that could be found in the reports.  This probably means they use SMB 1 and not the latest SMB 2.1 (from Microsoft) which might speed it up even more. On the other hand, SPECsfs2008 clearly states that it uses NFSv3.  Unclear to us whether NFSv4 would help speed NFS throughput per disk up or not.

As always we welcome any recommendations for improvement of our SPECsfs2008 analysis.  For the discriminating storage analyst or for anyone who wants to learn more we now include a top 30 version of these and all our other charts plus further refined performance 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 September 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 storage system features and performance please take the time to examine our NAS Buying Guide available for purchase from our website.]

~~~~

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

Data ONTAP 8.1 enhancements for Cluster-Mode data

Enhanced data mobility

Now Cluster-Mode data can be non-disruptively migrated from one cluster storage system to another.  This is especially useful for technology refresh as an older storage system can now be depleted of data, removed from the cluster and a new storage system inserted to replace it. Another use case is to load balance the IO activity across the clustered storage systems.

The unit of data mobility is a Cluster-Mode volume and all of its attributes are moved along with the data.  That is thin provisioning, deduplication, and data compression are all moved right along with the data from one storage system to another.  As you may recall, compression and deduplication are done only within a volume.  As such, the compression and deduplication dictionaries are also maintained at the volume level.  So when the volume data is moved so is its deduplication/compression dictionary.  In addition, SnapMirror when active for Cluster-Mode volumes is also cognizant of data mobility and where data is moved, all its replication services move right along with it.

Data mobility has some unique benefits for common applications and has been qualified and tested with applications such as VMware, Microsoft Exchange, Sharepoint. SQL server and Hyper-V, Oracle Database, IBM ClearCase and Siemens PLM.  For instance, when using Cluster-Mode volumes with Microsoft Exchange 2010, one can move mailbox database volumes much easier across the cluster than doing this with Exchange facilities alone.

Also, NetApp’s secure, multi-tenancy capability, can now span clustered servers. As such, a service provider can isolate a number of volumes from other client activity and move any of those volumes to other storage systems in the cluster while still preserving isolation and security.

All this is accomplished by synchronously replicating the volume meta-data across all the storage systems in the cluster.   That way all the systems have visibility to all the Cluster-Mode volumes in the cluster.

Scaleable SAN

For block storage, Data ONTAP 8.1 now offers a brand new capability, block protocol access (FC, FCoE, and iSCSI) to a scaleable storage cluster. In this case, LUN volumes can be migrated from one NetApp storage system in the cluster to another without operations intervention.

Cluster characteristics

Cluster interconnect is 10GBE.  File data clusters can span 24 storage systems, SAN data clusters are currently limited to 4 storage systems, but this is all that has been tested at this time and will likely be expanded as needed.

Moreover, clusters can have any type of NetApp storage system currently supporting Data ONTAP 8 which means from entry level FAS2040 to high-end FAS6280, including systems with and without Flash Cache, SSDs, SAS and/or SATA drives.  In addition, Cluster-Mode can also be used for V-Series virtualized external storage providing all the same capabilities to heterogeneous storage pools behind the V-Series controllers.

Other Cluster-Mode enhancements

In addition, antivirus software can be run on the storage nodes.  Also Cluster-Mode now supports NFSv4 and NFSv4.1 including pNFS. Further, Data ONTAP 8.1 provides support for Microsoft’s Server Message Block (SMB) 2.1 for CIFS.

Data ONTAP 8.1 enhancments for 7-Mode data

For 7-Mode data, Data ONTAP 8.1 provides a superset of all prior ONTAP 7-Mode capabilities, which now includes SnapLock®, IPv6 support, DataMotion™ enhancements and support for SAS storage in a MetroCluster™.  The belief is that 7-Mode customers running on old generations of Data ONTAP can now move to the latest version with no restrictions.

Announcement Significance

Cluster-Mode has had a long gestation at NetApp but one can finally see where they are heading with the technology. The new data mobility should appeal to a much larger portion of their customer base than previously.  With prior versions of Cluster-Mode, mainly media and entertainment and HPC/scientific computing were the main users. With this new release many more enterprise customers should be interested.

As for 7-Mode, the time is nigh to move to the latest code base, there doesn’t seem to be any rational reason to linger on old code anymore.  In addition, it’s unclear what if any 7-Mode capabilities are not available to Cluster-Mode data, but whatever the list was it’s getting much smaller with each new release.

[This storage announcement dispatch was originally sent out to our newsletter subscribers in September 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.]

~~~~

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

SPC-2*results

We start our discussion with the top 10 $/MBPS a sort of price performance metric for SPC-2 throughput results.

SCISPC110822-001 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 1 Top 10 SPC-2* $/MBPS

Lower is better on the $/MBPS results.  Here we can see the new Fujitsu DX80 S2 coming in at #7 on this chart. The series 2 version of the DX80 used 10Krpm, 2.5” disk drives for their submission and although it doubled the MBPS of the previous generation DX80 system it didn’t beat it in price performance, measured here as $/MBPS.   But the first generation DX80 subsystem was running 15Krpm disks with a much smaller (read less expensive) system and as such, did slightly better coming in at #5 on this price performance metric.

Next let’s turn to storage performance versus number of disks.

SCISPC110822-002 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 2 Scatter plot LDQ MBPS vs. number of drives

This is our first attempt at showing SPC-2 throughput vs. number of drives. As most of you that follow my prior dispatches or (RayOnStorage) blog know, there has been a lot of discussion around the theme that IO benchmark results are mostly determined by number of disk drives used in the submission.  We have started to product performance vs. drive spindles scatter plots to ascertain the truth to this viewpoint.

Above, we show the correlation between number of disk drives and SPC-2’s Large Database Query (LDQ) MBPS achieved, which at an R**2 of ~0.41, does not help their case.  We chose LDQ because it had the best correlation of the workloads used in SPC-2 that includes video on demand and large file processing along with LDQ.

Given what we see here, we would have to conclude that the number of disk spindles does not entirely drive SPC-2 MBPS results.

SPC-1 results

Although there has been no new data in SPC-1 results we thought it worthwhile to show another scatter plot.

SCISPC110822-003 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 3 SPC-1 IOPS vs. drive spindles scatter plot

In contrast to SPC-2 results abave, the same type of scatter plot for SPC-1 shows a pretty decent correlation of ~0.93, indicating that the number of disk spindles drives much of the SPC-1 IOPS results seen in the testing.  Why SPC-1 IOPS would be determined by drive counts but SPC-2 would not is worth considering.

One thought is that SPC-1 IOPS might be better distributed across the drives in a system while SPC-2 workloads are not.  This could be an easy explanation to the results seen in the scatter plots shown here.

Significance

It’s good to see some recent SPC-2 submissions but we still would like to see more.  Similarly, SPC-1 could use a few new submissions now that all the recent spate of acquisitions have worked themselves out.  But it is summer, perhaps by the fall SPC-1 activity will pick up.

We always learn something from looking at benchmark results.  The scatter plot analysis of SPC-2 and SPC-1 shows how those two workloads differ at least with respect to the impact of drive counts.  It’s too bad that not all SPC-1 submissions are required to also provide an SPC-2 submission and vice versa.  Although my vendor friends would probably not be happy to do this considering the cost and time involved to do both benchmarks.  In our opinion, just like IOPS and LRT are two distinct, complementary dimensions of storage horsepower, throughput also provides another unique, orthogonal measure of IO performance. But taking all three together can provide a more well rounded assessment of storage system capabilities.

As always, any suggestions on how to improve our performance analyses are welcomed.  Additionally, if you are interested in more details, we now provide a fuller version of all these charts in SCI’s SAN Storage Buying Guide which can be purchased separately from our website[1].

[This performance dispatch was originally sent out to our newsletter subscribers in August 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 SAN storage system features and performance please take the time to examine our SAN Storage Buying Guide available for purchase from our website.]

—–

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

EMC recently announced a new version of their disk library for mainframe environments the DLm 6000 with new backend storage capabilities.

DLm6000

For the first time last year EMC rolled out a new deduplicition capability for their DLm product using a Data Domain deduplication appliance.  At that time the DLMm960 product used the Celerra NS960 and the then current Data Domain DD880 appliance.

The new DLm6000 upgrades the backend hardware to the latest VNX 7500 and the newest Data Domain DD890 and upgrades the frontend with new Bustec hardware.  The new system supports up to 2GB/s of throughput and scales to 5.7PB of logical capacity (after deduplication).  With all the new hardware capabilities EMC believes their new system is 2X faster than the nearest mainframe VTL competitor.

The new Bustec frontent supports up to 6 virtual tape engines (VTEs) emulating IBM tape drives.  Tape volumes are stored as files and are accessible across all tape engines.  VTEs handle tape image creation and writes these images to the backend storage.

Recall that with the previous generation DLm960 one could select between alternative backend targets for virtual tape activity, i.e., normal disk storage for high throughput, non-deduplicatable data or Data Domain storage for highly deduplicatable data such as full backups.

In addition to the hardware upgrades, the new DLm6000 centralizes operations utilizing the z/OS console.  Also the DLm6000 now provides finer grained replication available now at the file system vs. disk target level.

Furthermore, EMC also offers a Data Domain systems using Bustec VTE’s, in a gateway configuration, for those customers just needing backup services.  In these configurations, the backend Data Domain appliance can be shared between mainframe and open system environments.

Tape use in mainframe environments

As most mainframe environments matured in an environment where tape was by far the most economical storage alternative, tape continues to be heavily used in this space.  Not only is tape employed as a backup medium, but it also plays a critical role in secondary data storage, such as for DFHSM (hierarchical storage manager) as a tier 2 or 3 backing store for disk datasets.  Also in some situations, tape can be exploited as primary storage such as for temporary workspaces used in large dataset calculations and processing in this sense a pre-cursor to today’s big data analytics, using batch processing.

For example, one customer using the DLm6000 for HSM backing store reduced wait times to 1 second on average from the 90 seconds they saw previously with tape infrastructure and as such, provided $450K in annual personnel savings.  The other major concern for this customer was DR readiness.  With the new DLm6000 data replication services they were able to verify that data was available at their secondary sites.

Announcement significance

There is no doubt that in open systems, traditional tape markets are giving way to disk-based solutions, relegating tape to more long-term storage archive applications.  Mainframes, due to their long history, probably use tape more extensively and efficiently than open systems and remain one of the last bastions of heavy tape use outside data archive.

With the introduction last year of DLm960 and this year’s DLm6000, EMC is starting to attack this remaining stronghold of tape use.  Nevertheless, we predict mainframes will not migrate quickly from a relatively tape intensive environment to a disk based solution.

For instance, mainframe systems that combine the use of disk and physical tape have been around from StorageTek since the late-90’s and from IBM for about 3 years.  These products primarily cache data to disk and then migrate it to tape.  As such, systems like these can provide an alternate, hybrid disk accelerated tape path for traditional tape mainframe workloads.  Further, many customers have already invested in tape automation and when doing so, find that tape storage economics can be hard to beat.

On the other hand, EMC is a disk-only system, doing away with tape infrastructure altogether.  Moreover, EMC brings deduplication technology to this market, which the other hybrid disk and physical tape systems currently lack.  Deduplication turned out to be the key technology that caused open system backup to move from physical tape to disk based systems.  EMC is betting that deduplication will have a similar impact in the more traditional mainframe environment.  But mainframe tape use is much more varied than simply backups, which is why DLm6000 has two target backends.

However, at least one things for sure, mainframes are becoming a new battleground in the age-old conflict between disk and tape.  One that originally started there, over 40 years ago, with IBM’s own admission that tape was dead.

[This announcement summary dispatch was originally sent out to our newsletter subscribers in August 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.]

—-

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

Latest ESRP V3.0 results

We start our ESRP analysis with Exchange database and log access latency results.  Recall that this chart is sorted by database read latencies.

SCIESRP110726-001 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 1 Top ESRP database-read latencies

For Figure 1, the new submission for the Fujitsu Eternus JX40 came in as number one.  One possible reason for Eternus’s good response times is the use of lagged DAG for replication.  The other new additions to this chart since last time were the HP P2000 G3 iSCSI systems coming it at #5 and 6.  One can see a nice mixture of SAS (#1, 3, 4, 9 & 10), iSCSI (#5, 6 & 7) and FC systems (#2 & 8).  However, there doesn’t seem to be high correlation between latency and interface used in ESRP workloads.

There has been much controversy on my blog[3] about ESRP’s latency numbers.  Microsoft caps latencies at 20 msec., which means ESRP reports will never exceed that duration.  Nonetheless, I feel that good latency is still something to strive for regardless of any cutoff.  Others suggested ESRP latencies were now a pass-fail grade and should not be ranked.  Clearly, I disagree with this sentiment.

We now turn to normalized ESRP database transfers per second.

SCIESRP110726-002 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 2 Top 10 Database normalized transfers per second

There are five new submissions on this chart namely, the (#2 & 3) HP iSCSI systems and (#7, 9 & 10) HP E5700 systems at 120, 80 and 40 TB respectively.  Unclear what #1 HP P2000 SAS G3 MSA did that was so great but it’s a screamer in both normalized and un-normalized data base transfers/sec. But there were plenty of other systems using SAS both from HP as well as other vendors that didn’t show this level of performance.

Next we examine log playback performance.

SCIESRP110726-003 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 3 Top log playback times

This is one of my favorite ESRP metrics because:

• It’s independent of Exchange and ESRP configuration parameters, i.e., it uses similar 1MB log files for any and all storage subsystems
• It’s a mixture of log and database processing that’s done as fast as the storage system and server(s) can accomplish it.

There have been only two new submissions here (#1 & 2), the new HP P2000 iSCSI systems that used 10GBE interfaces.  However, all the HP P2000 storage systems performed well in log processing.

Another of my favorite ESRP reported metrics is DB backup activity. ESRP publishes two measurements on database backup activity 1) Data backed up as an average MB/sec/database and 2) Data backed up as an average MB/sec/server.  Neither backup metric is entirely useful from our perspective so we have computed a new one.  Specifically, we multiply the average MB/sec/database times the number of databases to calculate a total database backup in MB/sec.

SCIESRP110726-004 (c) 2011 Silverton Consulting, Inc., All Rights Reserved

Figure 4 Top 10 total database backup rates

Figure 5, summarizes our new metric and ranking the top 10 storage systems in total database backup rates.  One would think that more disk drives would provide better backup performance but that isn’t the case here.  The result with the most disk drives is #10 the Fujitsu Eternus storage with 144 drives and the one with the least disk drives is #2 Dell PowerEdge R510 with 12 disks.  The #1 HP E5700 120TB solution had 56 drives, close to average for all submissions in this 1K-to-5K mailbox category.

My main problem with ESRP currently reported backup measures is that they are somewhat arbitrary. The MB/sec/DB metric could vary all over the place depending on how many databases were in the Exchange configuration for the same storage configuration.  Similarly, the MB/sec/server measure could also substantially change depending on the number of servers used in your submission. On the other hand, the total database backup rate should give a truer accounting of the backup rate of the storage regardless of the server or database configurations.  Naturally, this is a matter of opinion.

Conclusions

Well in our second ESRP analysis on the 1K-to-5K mailbox category we had quite a lot to discuss.  We are constantly trying to better understand Exchange 2010 storage performance and our new chart on total database backup is one result of that analysis.

Nevertheless, if we only had three charts to depict Exchange performance as revealed by ESRP report they would be 1) average database transaction latency, 2) average log playback time, and 3) our new total database backup chart.  We believe that these three charts provide the best complete view of storage system performance for Exchange, showing storage response time, transaction processing speed and sequential throughput.

So, why report on normalized transactions per second? It does illustrate what can be accomplished with the right Exchange and storage environment but it’s not an exclusive storage performance metric.  Just compare the log playback and normalized transactions processing rankings to see the relatively low correlation between them. We believe such a low correlation indicates that Exchange configuration parameters can highly influence database transaction processing results.  This is not true for log processing results.

Finally, as discussed in prior dispatches, ESRP/Jetstress results seem destined to be difficult to compare but in our view, merit the effort.  As such, feel free to contact us with any constructive ideas on how to improve.  In that regard, our contact information can be found below or on our website at SilvertonConsulting.com.

[This performance dispatch was originally sent out to our newsletter subscribers in July 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 SAN storage system features and performance please take the time to examine our SAN Storage Buying Guide available for purchase from our website.]

—–

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

IBM recently announced a new version of their XIV storage subsystem, with higher performance, higher entry point and new SSD cache upgradeability.

New XIV storage

The new XIV gen 3 storage system has been redesigned to focus on applications that need high IO performance.  Enhancements were made to hardware and software functionality to increase both random and sequential IO throughput.

For instance, XIV gen 3 hardware changes include:

• Infiniband interconnect – The new node interconnect reduces backplane latency  (<1 microsecond) and increases inter-node bandwidth by 20X.
• More processing power – The new hardware supplies 60 hyper-threaded, Intel processing cores used for more aggressive caching and better system management.
• More cache – Gen 3 supports up to 360GB of cache, which should help IO performance.
• Faster FC – The new 8Gbps FC frontend ports should increase system IO bandwidth and speed up access to data.
• SAS backend and drives – Gen 3 supports 6Gbps SAS backend and 2TB nearline drives which should help sequential throughput.
• More iSCSI ports – The new hardware supports up to 22-GigE iSCSI ports for better throughput and configuration flexibility.
• Future-proofed hardware – Gen 3 ships with enough spare DIMM slots for an additional 48GB of memory, a spare CPU socket, and in the future, will support up to 7.5TB of SSD cache which should increase performance even more.

Although 90% of XIV code is similar between gen 2 and 3, there have been a few enhancements to better support the new hardware.  Specifically,

More aggressive data pre-fetch – which reads data into cache where it can be accessed faster than off disk.  All XIV modules participate in this distributed, pre-fetch of customer data.

• More space efficient – which increases the amount of data that can be stored in the same physical capacity for thinly provisioned LUNs and all storage snapshots – with particular improvement expected for VMware VMFS datastore LUNs.
• Asynchronous replication improvements – reduces host latency for heavy small block random IO workloads by up to 4X.
• Multi-system UI – which eases storage administration for multi-XIV environments.
• Mobile UI – will supply XIV near real-time performance analysis on mobile devices like the Apple iPad to take along to meetings and other interactions outside the normal office environment.
• vCenter plugin –which allows a VMware system administrator to configure, monitor and manage the storage without needing to understand or use XIV’s UI.

Most of the above changes have led to increased system IO performance, even without using SSDs. As a result, XIV gen 3

• Sequential bandwidth has increased to 10GB/sec, a 4X improvement,
• OLTP/Transactional performance has improved by up to 2X, and
• Read hit workloads can now attain 500K IOPs with IO latency under 170 microseconds.

With all the hardware advances in the new system, there is no interoperability between the older XIV gen 2 and gen 3 storage but they can replicate to one another. Moreover, XIV supports a data migration utility that can, with minimal disruption, be used to migrate data from gen 2 (or any FC storage) to gen 3 systems.

Announcement significance

Previously, XIV’s major weakness was IO performance.  With this announcement, it seems like gen 3 should put most of those concerns to rest.  However, we would need to see some storage benchmark results to validate their performance claims.

XIV has always been known for its ease of use.  With the introduction of a mobile UI and multi-system operations, they have made it even easier to use.  In fact, one can see some of the XIV UI capabilities being introduced into other IBM storage systems like the Storwize v7000 and DS8000 storage systems.

On the other hand, probably more intriguing is gen 3’s empty hardware slots.  When you add this to all the processing capabilities currently available, it tells us more software functionality is forthcoming.  It’s unclear what IBM plans to do with all that spare power, but we can certainly come up with a few ideas.

[This announcement summary dispatch was originally sent out to our newsletter subscribers in July 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.]

—-

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