Not so FAST: Responding to Hitachi and HP FAST Claims

As one might expect, this incredibly important and innovative feature from EMC is beginning to bring out the “Me-too,” “We had it first,” and the “You don’t need it because we have a better solution” crowds. Since the announcement of EMC’s FAST solution for Symmetrix V-Max and other EMC  storage arrays, Hitachi and HP have been claiming that they’ve had a similar capability since 2000.

Nice try.

When Hitachi or HP try to claim their data migration tools are comparable to FAST or claim that they’ve had automation since 2000 or so, I start by first defining FAST, as it appears that they misunderstand or ignore what FAST is all about: FAST (Fully Automated Storage Tiering) is an automated, policy-driven method of placing the right data on the right storage at the right time, allowing for unattended and quick response to rapidly changing business and application requirements. FAST does the research for you to uncover migration and target candidates, as opposed to other methods that require user input or effort. FAST is not a performance tuning or load balancing application. Other competent tools provide those capabilities.

What are Hitachi and HP Claiming?

• They’ve had automated storage tiering in their high-end arrays since 2000, and in 2004 they claim to have extended automated tiering to externally attached storage.
• Tiered Storage Manager “allows you to automatically and non-disruptively migrate data between tiers of storage while applications remain on line.”
• “You set performance goals and limits, and HP XP AutoLUN Software does the rest.”
• “The capabilities are built into the XP’s firmware and therefore native and are licensed key enabled.”
• DMX and V-Max require external software to read and write data across internal tiers of storage and move or copy data from the DMX to V-Max.
• Tiering by File is the way to go, not volume.
  

Reality Check

Hitachi and HP have two data movement applications: Hitachi Volume Migration and HP AutoLUN (same product – formerly called CruiseControl) and Tiered Storage Manager. Neither have FAST-like capabilities.

1. AutoLUN/Volume Migration is similar to Symmetrix Optimizer. It is a backend load balancing tool that has nothing to do with storage tiering or moving a volume to a higher performing tier as necessary. It was designed to eliminate hot spots on drives. There are key differences that make EMC Symmetrix Optimizer a better choice than HP AutoLUN.

   • One key differentiator is that Symmetrix Optimizer uses a more sophisticated algorithm to determine where to place the migrating volume. It selects available space not only by Raid type, but drive geometry as well. This assures that the volume will be migrated to the best available performing platter, and location on the platter. AutoLUN does not have this capability. It only looks for available space on cooler raid groups.

   • AutoLUN requires a user to reserve target volumes prior to a migration. This means that either user intervention is required prior to each migration, or a massive amount of capacity must be reserved on each tier so that an offending volume can be migrated to a similarly-performing target. Symmetrix Optimizer does not require reserved space because if none is available, it can perform a swap with a cold volume or hyper. And of course, all of this can be defined by user policy.

   • Symmetrix Optimizer does not have to move an entire volume as does AutoLUN. Optimizer can move a much smaller Hyper. This provides faster migrations with less impact on production applications.

   • Optimizer, as with any other hot spot remediation application, is not necessary for thin volumes as thin volumes have their data striped across all the physical drives in a pool, thereby avoiding the hot spot issue.

   • Yes, AutoLUN is automated as is Symmetrix Optimizer. But balancing the workload across backend directors has nothing to do with storage tiering or managing a specific service level agreement for a particular application, volume or device. AutoLUN, as well as Optimizer, is a coarse method of assuring that all backend components are being used effectively, and that there are no hot spots on a particular platter. AutoLUN is included in the microcode, but I’m not sure why HP would believe that’s important or why they think it is a differentiator? The same can be said for Symmetrix Optimizer, VLUN and FAST.

   Just to make sure that the point has been made, here is the Hitachi tag line for Volume Migration:
   
   “Volume Migration: Data migration tuning to maximize subsystem backend performance.”
   
   Note that it doesn’t say anything about storage tiering.

2. Tiered Storage Manager is best compared to Symmetrix VLUN technology, in that they both require user effort to effectively perform migrations across Raid and drive types – or tiers. And they are both based upon their respective migration engines (AutoLUN or Optimizer.) Although HP claims TSM is automated, only limited automation can be accomplished through its scripting – none exists via the GUI. HP might also claim that Tiered Storage Manager is FAST capable. Let’s look at some of the key differences.

• Performance-based migrations. HP will claim that TSM can use input from Performance Advisor (PA) to make a decision on which volumes to migrate. However, the input is limited to Average and Maximum Utilization for Array Groups (Raid group) and not volumes (this is stated in the HP TSM User’s Guide.) In addition, this summary information is only published to TSM from Performance Advisor on a weekly or monthly basis – user selectable. So once the user has found a heavily used array group from information that is at least a week old, he then has to go back to Performance Advisor to identify the best migration candidates (volumes) on that Raid group. The user also has to select the destination or target group. Some of this can be pre-configured, but I don’t believe that a customer would want to migrate a volume based upon week- or month-old performance data, especially if they are trying to manage application service level agreements. FAST will acquire and utilize performance data having much, much smaller granularity - let’s say near real-time.

• TSM is limited to 8 simultaneous migrations. Once the 8 volumes are migrated, another 8 can be moved. During migrations on the HP XP24000, we’ve measured significant performance impact on the volumes being migrated. TSM does not allow pacing, so it tries to migrate the volumes as quickly as it can, even at the cost of production applications. It’s also possible that if too many IOs are being generated to the migrating volumes (more than about 50,) the migration will fail (as noted in the User’s Guide.) Compare that to VLUN (not FAST.) VLUN in V-Max can migrate up to 128 volumes per engine, or about 40 in DMX4 with negligible impact on production. In addition, the user can set pacing so that the migration will have even less impact on production, but might take a bit longer.

• During migrations, the user must stop Continuous Access Journal (Hitachi Universal Replicator) the async distance replication application. VLUN and FAST can migrate volumes non-disruptively across EMC's array-based replication processes.

• TSM requires the HP CommandView Advanced Edition XP (Hitachi Device Manager) management platform that runs on an external management server, because TSM is a plug-in application to CommandView, So in this case, the customer must purchase CommandView, TSM, and a management server. FAST is array based and only requires Optimizer.

• Hitachi has published a paper on the performance implications of using TSM. Here is a brief excerpt:

"This suite of tests demonstrated that the migration of data volumes with Tiered Storage Manager software does affect the performance (or response time) of the application being migrated, and can also affect the performance of other applications that may be accessing data on either the source or the target array group.

While there are additional tests that these findings suggest, there are a number of observations worth noting:

- In each case there was some degradation of response time for the application while it was being migrated. In some cases there was a significant difference - application response times up to three of four times longer.

- The impact on application response time often showed a significant degradation at the final stage when the volume pointers are swapped and cache is invalidated. This degradation period has a short duration. Performance then improves as the cache starts to fill with ongoing data access."

So you can see that HP AutoLUN, Hitachi Volume Migration and Tiered Storage Manager have little in common with EMC's FAST. They are designed for backend load balancing or manually initiated migrations based upon week-old or greater performance data. Symmetrix FAST, on the other hand, is a real-time or near real-time fully automated migration tool that will put data on the right set of spindles or Flash when needed.

So once again, HP and Hitachi are scrambling for a response or trying to redirect the discussion.

Jerry Zeisler
EMC Technical Business Consultant
Posted to EMC|One on October 29, 2009
[Reprinted with permission]

but don't just take it from Jerry…

Like me, Jerry works for EMC. So of course, many of you will discount the above as merely another round of FUD and competitor bashing.

Fair enough.

But here's how Hitachi positioned TSM at a recent Web Seminar (click for full-sized version):

Any way you look at it, that isn't a description of AUTOMATED Tiered Storage (much less FULLY automated ).

technorati tags: EMC,Symmetrix,V-Max,FAST,Fully Automated Tiered Storage,Hitachi,Tiered Storage Manager,Hitachi Volume Migration,HP AutoLUN,Symmetrix Optimizer,tiered storage

I can't sit on the sidelines and let that post go unanswered.

Hu is asserting that scale-out cannot support the Virtual Server world, but he makes no real case to back this assertion up. More importantly, it appears to me that he still doesn't understand how V-Max scale-out works.

What's more, his attacks on “modular” appear to be nothing more than an obvious attempt to defend Hitachi's “monolithic” architecture. And those of us with longer memories will recognize that Hu's position is in fact a 180-degree reversal of the position HDS took against Symmetrix for the last decade or so – Hu rarely missed an opportunity to beat EMC up for the fixed-cabinet “monolithic” storage of Symm 5 and DMX 1&2. Hitachi even had slides (and a few You-Tube videos) attacking the “Symmetrix monolith”.

And here Hu is today defending the very monolithic approach that HDS so aggressively challenged not so long ago.

v-max: purpose-built for the virtual data center(tm)

Clearly, Hu is also trying to challenge EMC's positioning that the V-Max is specifically optimized to support the needs of virtual data center.

Hu’s assertion that scale-out can’t meet the needs is nothing more than uninformed (or defensive) FUD. That he's pointed his attack on the architecture serves a very real issue that Hitachi would rather we not discuss: configuring, allocating and managing storage devices in a VMware world with the USP-V/VM is outrageously complex. I have had the opportunity to present customers with the specific software optimizations that have gone into V-Max to simplify storage deployment in a VMware environment, and to compare and contrast V-Max's advantages to the USP-V. To a customer & prospect, everyone familiar with Hitachi has told me that I was far too kind in those comparisons – they collectively confirmed that V-Max's software functionality and storage management is light years ahead when it comes to streamlining the integration and support of virtual servers.

And the fact is that V-Max is still, at its heart, a Symmetrix. Any I/O can be serviced by any engine/director/front-end port. Workloads can be distributed across as many engines/directors/back-end ports and drives as necessary to meet the required workloads. All memory across all of the engines is available to every application, and/or memory can be dynamically partitioned to insulate applications from one another. With V-Max you can start small, with a single engine, and add more engines as your workloads expand (or as you add new ESX clusters, for example). You can non-disruptively rebalance those workloads across the larger engines, or you can choose to segregate workloads to a subset of the engines if you choose.

On one hand, the “directors” in V-Max are synonymous with the blade directors we used to use in prior generations – they work exactly the same, except in V-Max we combine front-end, back-end and memory into a single V-Max director, then combine two of these into a high-availability V-Max engine. As with all prior generations of Symmetrix, V-Max memory is a single global resource, and any engine/director/CPU/core can access any memory location on any engine/director in the Virtual Matrix for both cache and meta-data operations.

On the other hand, the result of clustering multiple independent HA engines to form a larger system that delivers a true scale-out architecture, provides most of the good things about scale UP, without the cost or space limitations of historical “monolithic” approaches.

scale-out for performance

As to the performance assertions of “scale up” that Hu makes, a single V-Max engine delivers more performance/throughput/IOPs than does the smallest DMX4-950. So indeed, with V-Max we have scaled “up” the basic building block of a Symmetrix, even as we have shifted from PowerPC to the Intel IA architecture. The differentiating factor is that now we can scale OUT these engines to create a system that can be incrementally expanded to support capacities and performance that FAR EXCEED the relatively minute capabilities of Hitachi’s monolithic enterprise array. In fact, V-Max can today address more than twice as much usable internal capacity as a USP-V; and it does that without sacrificing availability or performance (as does USP-V/VM whenever used to virtualize external storage).

More importantly, V-Max is now on the IA performance curve: future generations will leverage the performance and architectural improvements that Intel delivers on their roadmap. And as an aside, with Pat Gelsinger (ex-head of IA at Intel) now running EMC's storage platforms business, you can be assured we're already being pressured to accelerate our IA integration roadmap.

One of the important thing that the accelerating IA platform brings to the table is ever faster CPU cycles. These, in turn, allow Symmetrix software engineers to optimize and extend the cache pre-fetch optimizations to better handle the way that virtual machine consolidation has changed the IO path. Here again I disagree with Hu: far from becoming anything as coherent and (dare I say it) predictable as mainframe IO patterns, server consolidation truly presents near chaos of intermixed I/O requests over each HBA. Here again, V-Max has been tuned and optimized to dissect these IO streams and to extract the patterns that serve as hints to the powerful pre-fetch engine. As a result, V-Max continues the Symmetrix legacy of delivering remarkably high cache hit rates, even as workloads change and move across servers or clusters.

bottom line

I contend Hu’s assertions that scale-out cannot meet the needs of the Virtual Data Center (aka Private Cloud) are entirely self-serving; Hitachi does not have an answer to the scale-out architecture that is increasingly becoming prevalent across the entire market landscape, and thus Hu has no choice but to argue against it.

EMC, along with the rapidly expanding V-Max installed-base, respectfully disagree with Hu.

technorati tags: V-Max,EMC,Symmetrix,DMX,cache hit rates,scale-out,storage,architecture,Hitachi,HDS,Hu Yoshida,USP-V,USP-VM,scale-up,dynamic cache partitioning,pre-fetch,cache,caching,VMware,VMware Integration,virtual data center,private cloud,vdc

In no particular order, here are several namesakes of IBM's latest enterprise-class storage wanna-bee:

DS-8700 High-speed lockstitch sewing machine

Dongsen's website description for the DS-8700 starts off with "A new generation model designed for a quiet and smooth performance," a statement that pretty much confirms that this isn't your typical storage product. And not surprisingly, there's also a knock-off version of the DS-8700 from DASU, marketed as the DS8700 (without the hyphen – how clever).

Same exact model number, with just enough visual differentiation to sidestep any legal concerns… 
 

Servo DS 8700 G Digital-Super Graupner

The description of this clever little device could perhaps be mistaken as referencing a storage product, although it might be some sort of self-propelled device (extremely fast; high torque; robust gearbox; works with all gyro systems):

Upon closer inspection, you realize that it's a servo for controlling the tail rotor for model helicopters, and not a mobile storage platform.

dallas semiconductor ds8700

This one had me chuckling.

Like IBM's DS8000 series, the Dallas Semiconductor microprocessor programming adapter model DS8700 is (already) obsolete, though you can still find a few laying around in various second-hand stores (Google search here). I couldn't find a picture, but I bet if we all click on a few of the links in that search the component liquidators may well interpret these as a sudden resurgence of interest!

Kinda the inverse of insider trading, if you catch my drift

owa-DS 8700

This one is my favorite, though: "especially for fishermen and hunters," with a compelling description:

• Available in 2 sizes
• See-Thru window graphics. You can see out. No one can see in.
• Only the designs and sizes displayed are available. No customizing of these designs.
• Generally ships in 2 business days.
• Easy to install high quality graphics

Makes you want to run to the telephone to call your IBM rep and order up a bunch, don't it?

the ibm ds-8700

There's not a lot more to say about IBM's DS8700 announcement today. A routine upgrade of processors from P5+ to P6's, a small increase in cache memory, elimination of ESCON and a switch from  RIO-G to PCIe Gen 2 as the processor/IO interconnect, consolidation into a single platform that can be configured 2-way or 4-way (I'm not sure if that's a field upgradable option, though).

Yawn!

Sure, there are the usual claims of 50% improvement in total IOPS over the DS8300 Turbo and a 150% increase in back-end bandwidth (MB/s). And my personal favorite was the claims of "up to 2.5x performance for distinct applications."

Somehow I doubt that "distinct" is synonymous with "real world."
Or even "commonplace," for that matter.

But running the claimed numbers against observed performance in both OLTP and DSS workloads on the DS8300, and the DS8700 still falls far short of competing with the DMX4, much less V-Max.

And if indeed all this claimed performance is real, then why is the new DS8700 still limited to a maximum of only 1024 disk drives? And why no mention of Flash drives? You'd think IBM would be whooping it up if all this performance translated into meaningful value for Flash drive deployments! And the lack of any reference to Flash drive improvements over the DS8300 implies that the system is still limited to using the older STEC 73GB and 146GB drives, while DMX4 and V-Max both support the newer 200GB and 400GB drives.

CORRECTION: IBM did indeed mention flash drives in their launch materials (I just missed it). Still limited to the smaller capacity drives, as I noted above. And the 8700 is limited to a max of 256 SSDs, no RAID 6 or RAID 10. Meanwhile, V-Max supports 200GB and 400GB flash drives, has no practical limit on the number of (you can RPQ up to 2300 or so if you'd like), and you can use any RAID protection type you prefer.

Most interesting to me was that there has been no mention of anything FAST-like for the DS8K, nor any real evidence of integration of Flash into the platform at all. This silence while Master Scientist Barry Whyte is literally gushing over on his blog about the fact that the SVC is getting native SAS flash capacity to use as a local mirror to virtualized storage. Given the relative silence from DS8K land, perhaps IBM has finally shifted its focus (and its "blue dollars") away from the aged old DS8K line and over to the Intel-powered SVC?

I'll bet that's why BarryW is so obviously enthusiastic these days!

But if IBM is investing in Flash only for the SVC platform, then the DS8700 must be part of an exit strategy…a final little speed-bump to tide things over just one more tech refresh cycle.

Bor-ing.

(Probably not even interesting enough for the DASU gang to want to clone it!)

Happy Sewing!
 

technorati tags: IBM,DS8000,DS8K,DS8700,DS-8700,obsolete technology

After nearly 3 years of denial, we now have proof-positive that the IBM DS8300 has been unceremoniously removed from life support. I've been told of numerous prospects whose IBM account teams vehemently denied the impending introduction of the DS8700 during the last quarter, even as EMC account teams asserted (with confidence) that the DS8300 was indeed on its deathbed.

True to my prediction back in February, the DS8700 intro is now inarguably imminent…see for yourself with this Google search. And the word is that many customers actually received quotes for the new DS8700 over the past several weeks.

Pity those who were suckered into buying a DS8300 this year (remember, I tried to warn you!)

This time I'll not play the role of truth-in-marketing (as I did for the XIV intro), so you'd better buckle your seat belts for another round of Meaningless Marketing coming from Big Blue as they try to convince you that the aging and decrepit Sharkitecture has been resuscitated with the magic face cream of P6 processors and the life-giving breath of flash drives.

i sincerely doubt that any of these will even come close to overcoming the inherent shortcomings of that architecture, though. Already I'm seeing outlandish claims that the DS8700 has ASTONISHING improvements for "distinct" workloads – which probably means they found some benchmark that looks good, even as the non-distinct workloads realize little or no new value.

Word is the DS8700 is part of next Tuesday's set of weekly IBM announcements, so it should be a fun couple of weeks in the blogosphere sifting through the FUD and marketing misrepresentations.

Cue TonyP!

technorati tags: IBM,XIV,DS8700,DS8300,Shark

You've got to admit that the lyrics of the Stealer's Wheel hit have an uncanny resonance with the FCoE misinformation being promoted by
you-know-Hu.

I hereby refer my readership to
Storagezilla's latest post on FCoE. 

technorati tags: FCoE,FUD

Although Chris Mellor scooped me with his second-hand coverage of the event (Chris leveraged a report put out by Aaron Rakers to customers of Stifel Nicolaus Equity Research for his story), I thought I'd share the session with my readers first-hand.

So, if you have an hour or so, pop over to the IR landing page on EMC.com and click the banner link to the recorded webcast (or go directly to the webcast hosting site). But take notice – this webcast will be available on-line only through October 21, 2009.

I also included the net result of the FAST demo that Chad Sakac presented at VMworld 2009 in yesterday's presentation. If you'd like to see the complete demonstration, it's available on YouTube here:

I have also received a few follow-up questions from the event; I'll answer several of them after the break…
 

fast follow-up in a flash

So here are a few questions that came in to EMC IR after yesterday's Tech Talk.

• I want to know more about alternative architectures which Barry mentioned at the end. Does it make sense to put EFDs in the server (next to the bus) instead of the SAN. Doesn't the SAN become a bottleneck with a traditional disk array configuration?
  
  As I have long asserted here in my blog, I am convinced that we will see Flash (and other persistent solid-state storage) play an ever increasing role all through the I/O and storage pipeline. We'll see Flash in the servers, in the network and in external storage arrays, both as persistent target storage and as cache devices. In fact, we're already seeing multiple tiers of Flash deployed for different use cases within a single I/O stack, each supporting different use cases or SLA policies…
  
  As to the common misperception that SAN overheads creates an insurmountable performance bottleneck, I merely point out that intelligent external storage arrays have been deploying the much-faster SDRAM as a cache buffer for slow disks for nearly 2 decades. This even while it is inarguable that the same SDRAM could be more quickly accessed were it in the servers instead. The benefits of storage consolidation and centralized BC/DR will continue to drive IT to leverage external storage, and of course the new performance capabilities of solid-state will ensure that external storage will evolve to maximize the benefits as well.

• I have some more technical questions about if the replaced drives (those replaced with Flash) are ones that were short-stroked. Does the same story on flash work if there are not short-stroked drives in the array?
  
  No, in this workload, the fibre channel disk drives were not short-stroked. Instead, the access density of the drives was clearly higher than what these HDDs could support with optimal response times. And in fact, had they actually been short-stroked, they still wouldn't have met the response times delivered by FAST (despite what the wide-striping consortia would have you believe).

• Following Flash and FAST appear to be the places EMC is driving competitive advantage. I would like a presentation on data warehousing as these new products may open that market to you in a way it was not open before.
  
  Yes, indeed – deploying FAST with Flash will undoubtedly change the face of not only on-line transaction processing, but data warehouse as well. Especially when we get to the sub-LUN version of FAST, where the hot blocks can be pre-staged into Flash. At that point, it should be less important that database indexes are optimized around the intended queries, because the traditional disk latencies will play an ever less significant role in query latency.

keep 'em coming

I'll be happy to answer any other questions on the subjects of Flash and FAST that I can - even those from the EMC competitor who continues to censor my comments on their blogs (you know HU you are).

So, ask away!

Oh, and here's the link to Chris' article on yesterday's webinar, in case you missed it.

Another original post by the storage anarchist.

technorati tags: EMC,EFD,FAST,V-Max,Symmetrix,Enterprise Flash Drives,Flash Drives,SATA,Solid State Storage,Fully Automated Storage Tiering

What better time to announce the No-Charge Symmetrix Migrator Package?

Effective this month, this new package provides both current and new V-Max and DMX customers with free licenses to three powerful storage migration utilities:

1. Open Migrator, for host-based migrations. Runs on most popular operating systems and server cluster software. OM can mount the new target LUN(s), copy data (on another array or from within the platform itself) in the background, and it will mirror writes to both old and new storage during the migration;
2. SRDF/DM (Data Mobility), Symmetrix-based replication frequently used to make a mirror of current Symmetrix volumes onto a new array before swapping the hosts over to the new storage. Used by customers for more than a decade to effect both Open Systems and Mainframe data migrations;
3. Open Replicator/LM (Live Migration), array-based "hot pull" migration for Open Systems hosts. Simply insert the new Symmetrix into the SAN in between the hosts and their current storage array(s). LUNs are copied off of the old storage in the background, while any host reads or writes are moved to the head of the queue for maximum efficiency. Works to migrate data into a Symmetrix from virtually any SAN-based storage platform, and it can even move a small existing LUN into a larger one in the process.

As with the recently announced FREE status for Symmetrix Virtual Provisioning, this new package comes with only one string: you have to own or purchase a Symmetrix DMX or V-Max. Other than that it is truly no-additional-charge. Customers can take advantage of these new tools immediately upon arrival of their system (or license keys), without concern for the amount of capacity they want to move (or Virtually Provision).

Yet another way that EMC is helping to reduce the TCO and expand the use cases of Symmetrix, the enterprise storage market share leader for nearly 20 years.

Enjoy!

technorati tags: EMC,Symmetrix,V-Max,DMX,Virtual Provisioning,Symmetrix Migrator Package,FREE,No-Charge,Open Replicator/LM,Open Migrator,SRDF/DM,data migration,tech refresh

As do most of you: Dick literally founded our industry.

I learned of Dick’s passing after having sat tearfully through several hours of poignant memorial for Senator Ted Kennedy on the television. Even though their politics surely have the two latest patrons of heaven's Irish pub seated at opposite ends of the bar, rest assured they’ve hoisted a pint in honor of each other’s contributions to the world they have left in our care.

To the both of them, a traditional Irish toast:

May your glass be ever full.
May the roof over your head be always strong.
And may you be in heaven
half an hour before the devil knows you're dead.

technorati tags: Richard Egan, EMC, Edward Kennedy

Drinking our own champagne.

This introductory episode outlines the quest for the VDCotF, and it stars none other than Symmetrix V-Max (and some guy from IT who seems intent in blocking our view).

Join me on this journey, if you will:

And yes, that's the same V-Max that starred in my "bootleg" video a few months ago.

Another inspiring perspective from the storage anarchist.

technorati tags: EMC,Symmetrix,V-Max,Virtualization,IT Infrastructure,VMware,VDI,Virtual Desktop Infrastructure,View,vSphere,cloud,private cloud,Virtual Storage

A couple of weeks ago I ran across this TED video where Mike Rowe of the Dirty Jobs show on The Discovery Channel relates a story about his unexpected experience working as a sheep farmer. If you're a fan of the show, you know that most of the Dirty Jobs that Mike undertakes also require a healthy dose of courage – Mike's day job is essentially to live Dave's metaphor day in and day out.

But this video caught my attention because where Dave focused on the challenge of the task, Mike's focus is on his learnings from completing it.

The entire segment is about 20 minutes long, but you'll get most of the point by about the half-way point.

WARNING: while this video doesn't include actual on-the-job footage, it is quite graphic. 
If you are at all squeamish about animal husbandry, I recommend that you just skip it.

Click here if you do not see the video above

DISCLAMER: I have never used a knife on live cattle or sheep

That said, and with all due respect to Dave, I think the shepherd's job requires just a teensy bit more intestinal fortitude. Don't you?.

For your convenience, here are links to the Wikipedia pages on Anagnorisis and Paripeteia
 

This is yet another thought-provoking perspective from the storage anarchist!
 

technorati tags: Dave Hicks,How to Castrate a Bull,Mike Rowe,Dirty Jobs,The Discovery Channel,castrate a lamb,anagnorisis,paripeteia,courage,fortitude,intrepidity,NTAP,NetApp,TED

Although Compellent's slow-and-steady growth isn't all that surprising, I for one was somewhat taken aback by the relatively poor performance of 3PAR. Especially given all the bluster and bravado that routinely comes our way from both David Scott (3PAR's president) and Marc Farley (3PAR's resident cheerleader cum rapper cum blogger) – listening to them in blissful ignorance of the facts would have you believe they're a lot bigger and more relevant than they really are. In fact, it strikes me that there's a disproportionate amount of noise coming from a storage company that still hasn't found its way onto the exponential growth curve after more than 10 years in business.

Add in the news that COPAN is struggling for relevance and ONSTor has been bought for pennies on the dollar, and the emerging estimates that Pillar Data is operating somewhere around the $44M annual revenue mark, and you get a pretty clear picture of chaos down at this end of the market.

read on for more perspectives…

struggling for relevance?

Knowing full well that I'm going to be on the receiving end of a gentlemanly wet noodle attack from Farley (he'll probably claim I'm so old I can only gum wet noodles anyway), I'll assert that indeed 3PAR is rapidly losing its differentiation. In fact, over the past several years, it seems that 3PAR has been jumping from one differentiator to the next, in hopes of staying dry as it crosses the raging river of competition.

And they're running out of rocks while the river is rapidly rising.

I recall several years back (when Farley was still at EqualLogic) customers came to EMC World and Customer Council demanding that we improve the time-to-first I/O for Symmetrix. It seemed then that 3PARs simplicity and speed of storage allocation was the differentiator that was attracting customers. And EMC listened: back then it took over an hour to allocate a terabyte of storage on a DMX; today it takes under 9 minutes on DMX3 and DMX4, and its even faster on the V-Max. And the Symmetrix Management Console (SMC) GUI has totally revitalized the process of operating and provisioning a Symmetrix array, with automation, templates that simplify everything while significantly reducing the opportunity for errors.

Next it was Thin Provisioning. Except to 3PAR's chagrin, it turned out not to really be about "thin" at all – customers made it clear that while they liked the idea of being able to over-provision their capacity and allocate-on-demand, they were scared as hell about the consequences of not having enough capacity – even if they held capacity in reserve. No, what they really wanted (for their Tier 1 storage at least) was the ability to provision storage without having to deal with physicals and hypers nor be limited to allocating LUNs in multiples of a standard hyper sizes. If an application owner requested 38GB, they wanted to be able to give them 38GB and not one kilobyte more.

And again EMC listened, with the introduction of Virtual Provisioning on DMX in 1Q'08. And in fact, Hitachi delivered Dynamic Provisioning on the USP-V just about 9 months earlier – combined, 3PAR no longer had a unique value proposition for the enterprise customer in thin provisioning.

And so the 3PAR song changed again – first to attacking Hitachi and EMC on their chunk size (David Scott is attributed with coining the term "chubby" provisioning last year), and when that didn't afford any traction they moved on to focusing on wide striping – only to find out that both Hitachi and Symmetrix' implementations afforded customers the same benefits. Wide striping was 3PAR's answer to Flash Drives too for a while (shades of IBM's "let them use TAPE!"), but eventually the marketing guy spilled the beans last October that indeed 3PAR would be adding EFDs to their lineup. Still hasn't happened, but thankfully that pretty much shut down Farley's "nobody needs flash" rap sessions.

Then along came V-Max, followed almost immediately with the sudden reorganization of the 3PARs technical leadership. As Chris Mellor chides, purely coincidental .

Still…it's interesting to see how all this falls out on that chart that Chris used. I've reproduced it here (using approximated data) with annotations on the timeline:

I'll let you draw your own conclusions as to cause and effect.

And now we have Farley's latest grasp-for-straws-of-differentiation: Reservationless (sic). As if being able to ensure that the most important applications on your array will always have the storage they require was a bad thing . 

What Farley (and others) seem to overlook is that when you have an array that can support 2PB of usable capacity made up of any mixture of EFD, FC and SATA capacities and speeds, you will almost by definition be supporting multiple different classes of applications on that platform. As such, it is an imperative that you be able to logically separate applications and their resources; Marc would have readers believe that defining separate pools for Production vs. Development was a bad thing. Symmetrix customers thank us for not listening to Marc – Dynamic Cache Partitioning, Priority Controls, Storage Groups and Virtual Provisioning Pools all serve to enable customers to safely maximize the value of consolidation and to protect their most important applications, even if a lesser application or a software development bug runs away uncontrollably consuming all available storage.

Unlike with restaurants, customers just can't take the chance that there won't be any seats available when they need more storage – that's why Symmetrix Virtual Provisioning allows for capacity to be both assigned to specific pools and to be pre-allocated for those applications that just can't wait. And while reservations aren't required, they truly are advised – especially for those applications that want need the very best.

That said, as EMC begins rolling out FAST (Fully Automated Storage Tiering) across all of its product lines later this year, I am sure the 3PAR choir will jump off the Reservationless rock and come up with a new song. But they are already playing from behind, lacking support for Enterprise Flash Drives in the first place, so I'm sure it will be interesting…you gotta give Farley one thing, he's damn creative!

a couple more perspectives

The other thing that struck me from Chris' original chart was that the scale made it appear as if  the battle is happening down in the $10-100M quarterly revenue range. With no disrespect to the small companies struggling to make their way to the big time, I believe it is more relevant to measure their progress not against each other, but in how they stack up against the larger vendors in the market.

Unfortunately, most of the larger companies don't break out the revenues of individual product lines, and I don't think IDC publishes all their data publicly. But while I cannot post actual product-line revenues, I can adjust the scale to reflect what it would have to be if we were to plot the major storage platforms' quarterly revenues alongside these four, comparing them to say, CLARiiON, DS8000, EVA, Symmetrix, and/or USP-V/XP/9990v, it would look something like this:

And just to keep things in grand perspective, if we scaled the chart out further to reflect the total external disk storage revenues of  the likes of EMC, Dell, HP, IBM, and/or NetApp, the chart would look something like this:

Clearly, the exit strategy for Compellent and 3PAR isn't to unseat the market leaders !

But if their strategy is to follow the footsteps of EqualLogic and Data Domain, my humble opinion is they are going to have to find the trail that leads to exponential growth – and fast. Given their rapidly fading product differentiation and their flat-to-declining recent performance, their long-term viability has got to be a concern for customers and shareholders alike. Especially in light of the apparent fate of the likes of ONSTor and COPAN.

And if they DO get bought out, I'm sure that their customers will be concerned that the new owners will protect their investment, as both EMC and DELL have done with DD & EL. That isn't easy, nor is it assured – only time will tell.

Well, that's my own personal opinion – I'm sure some of my readers will offer their own in response. As always, courteous comments will not be censored.
 

This is another insightful perspective from the storage anarchist.
 

technorati tags: 3PAR,Compellent,EqualLogic,Data Domain,Chris Mellor,COPAN,ONSTor,thin provisioning,wide striping,EFDs,Flash Drives,ease of use,Symmetrix,DMX,DMX3,DMX4,V-Max,Virtual Matrix,Dynamic Provisioning,Virtual Provisioning,DELL