NEWS & NOTES FROM THE THOUGHT LEADERS OF DATA STORAGE

Flash and Trash? Not so Fast

Mar

07

2010

It is said that world travel gives one perspective not otherwise achieved, seeing customers and the industry from a global view. So, on my current trip to South Africa and the UK, I took the time to catch up on my reading, in particular some thoughts on ‘Flash and Trash’ – i.e. the idea that the use of only two media – SSD and SATA – is going to dominate the future of storage. My take – not so fast.

Count me as Niels Bohr, when he said “Prediction is very difficult, especially about the future.” While the flash & trash thesis is interesting, I believe there is and always will be a place for high-speed, non-ATA (e.g. FC, SAS) media in both the cloud and internal datacenter. Having said that, I do understand why the proponents of flash & trash think as they do – because they don’t see the trend towards 2.5″ drives in enterprise, or perhaps don’t want to see it. This is, in my opinion, the trend to watch in storage.

Having said that, it’s vitally important to understand that enclosing and controlling 2.5″ drives is far different than our long-time friend, the 3.5″ drive. Nearly all of the current arrays, both small and large, cannot handle smaller drives due to their inherent speed and vibration. A 2.5″ 15K drive is a very innovative piece of electro-mechanical work, but most array vendors want nothing to do with it.

Xiotech, however, not only welcomes the 2.5″ FC/SAS drive technology but embraces it. I personally believe that in the near future – 2 years at most – we, as a company, will sell more 2.5″ datapacs than 3.5″. The work being done by the drive industry for use in high-speed, high-capacity 2.5″ datapacs is nothing short of outstanding.

There is little doubt that a datapac of 2.5″ drives – 20 per pac, 40 per 3U ISE – will form the large ‘middle’ for enterprise storage. While is it certainly convenient to believe the world will settle into flash & trash for storage, it is also certainly ignoring the very real factor of the datapac and its evolution into the world of very small, very fast drives. We will also likely see the 10,000 SPC-1 IOPS/3U (over a surface of tens of TB) barrier broken with this evolution.

Just as the 14″ drive gave way to 8″, and to 5.25″, and to 3.5″, we are now in the 2.5″ era for rotating storage. While flash has its place (small-capacity random-read-dominant workloads) and trash also has its place (deep-capacity, low-performance where data integrity is not important – since ATA technology does not support the ANSI Data Integrity Field standard) there is no denying the importance of the continued evolution of rotating magnetic storage.

Shortly, I will travel back to the States and home. It’s always wonderful to come home, of course, but looking at the world through a larger lens – and seeing the common thread worldwide for the need for efficient storage – is a great experience. If you get the chance, I urge you to do the same.

(See http://www.quotationspage.com/quotes/Niels_Bohr/ for more Bohr quotes – a good read)

Nova Networks Joins Xiotech Partner Program

Mar

04

2010

We announced this morning that Nova Networks has joined our Xcel Channel Partner Program. Through this relationship, Nova will re-sell our patented Intelligent Storage Element (ISE™) technology to various organizations.

For more information, please visit: http://xiotech.com/press-release.php?id=175.

What Price Storage?

Mar

02

2010

Chargeback models are becoming a common requirement for modern information technology (IT) operations.  As businesses have sought to not only control their costs for existing IT infrastructure but to understand them for future initiative planning, the model of IT operations providing hardware is giving way to that of them providing a service.  In this model, the tools of engagement are contracts rather than datasheets, service levels rather than hardware specifications.

Many IT organizations within existing companies are doing very well to adapt to this change, and those that aren’t are proving a fertile ground for takeover by managed service providers. But IT organizations are having to fight continual and unnecessary battles as they attempt to match their customers’ new-style demands with their vendors’ old-style offerings.

Chief amongst these in the storage industry is the habit by storage vendors to provide costs in terms of raw terabytes (TB).  This is advantageous to storage vendors because it allows them to avoid talking about the “hidden” requirements for storage.  Commonly, these include:

  • Drives withheld from use as they are required for sparing.  This is usually 10-15% of the total number of drives (for example, 2 drives per shelf of 16)
  • Storage used for RAID protection.  This is usually 14-50% of storage depending on the RAID level
  • Maximum suggested utilization of available storage.  This is usually around 70%

And there can be additional items such as space reserved for snapshots, asynchronous replication queues and the like.  Taking the best-case of each of the above, and remembering that the above numbers stack, this puts usable TB at 54% of raw TB (the worst, but not unusual, case puts you at less than 30%).  In other words, a purchase of 100TB of raw storage results in 54TB of usable storage.  Or to put it another way, if you are purchasing based on cost per raw TB, then chances are your storage is at least twice as expensive as you thought it was.

It is not so surprising that most storage vendors participate in the practice of quoting the far better-looking raw numbers and ignoring the real costs, but it is very surprising that so many customers allow them to do so.  If you’re looking to buy storage you should ask the following questions of your vendor:

  • How much space will I be able to present to my customers after sparing, RAID protection, snapshot reserve and any other non-chargeable storage use?
  • How full, in terms of a percentage, can my storage become before your best practices recommend that I purchase more?
  • What is the expected performance, in SPC-1 IOPS, of the proposed configuration?
  • What are the hardware and software maintenance and support costs I will need to pay over the 5-year lifetime of my storage?

With straight answers to the above questions, you will have two metrics, cost per usable TB and cost per IOPS, which you can use as a basis for comparing different vendors’ products to understand what you are getting for your money.  Additional metrics, such as capacity per U (unit of rack space) and performance per U may also be interesting for space-constrained environments, and GB per Watt and IOPS per Watt are useful for power-constrained environments.

Using cost per usable TB and cost per IOPS will enable far more accurate product comparisons than cost per raw TB, and provide you with the ability to build a true service-based model for your customers.  And if you are serious about a service-based approach, one additional question will allow you to understand if your storage vendor can really provide you with a scalable storage model or just a bunch of products:

  • How would the values of usable capacity, expected IOPS and total cost change if I were to add another 50% of raw capacity to the proposed solution?

If your vendor’s answer isn’t, “You’ll get 50% more usable capacity and performance, and it will cost you 50% more,” then you need to consider how you can provide a single service cost to your customers depending on the capacity and performance that they require rather than the specifics of the particular hardware solution which you are purchasing at that point in time.

There is little doubt that your customer requires cost per usable TB and cost per IOPS metrics to be able to manage their existing storage and project the cost of their future storage requirements.  The question is: what is your storage vendor hiding by not providing them to you?

Xiotech Launches Fellows Program, Names Inaugural Class

Feb

19

2010

Yesterday we launched our Xiotech Fellows program and named our inaugural class of Fellows.  As President and CEO Alan Atkinson said, “We at Xiotech are blessed to have an incredibly deep roster of storage ‘rock stars’ who have a collective body of accomplishments and patents that rivals anyone else in the industry.”

Our Fellows are:

  • Richard Lary, Corporate Fellow – With 40 years of industry experience and 32 patents, Lary’s legendary influence spans from the creation of VAX and the Digital Storage Architecture at DEC to having been a vital consultant to many corporations.
  • Rob Peglar, Senior Fellow – With 32 years of experience, Peglar is one of the industry’s most seasoned and recognized experts on a host of storage topics and the current treasurer of SNIA.
  • Ken Bates, Fellow – With 35 years of experience, Bates is a performance guru and was a guiding force in the development of the gold standard in performance testing – the SPC benchmark.
  • Todd Burkey, Fellow – With 33 years of experience, Burkey has helped to drive key Xiotech innovations for its industry leading reliability around predictive monitoring, telemetry automation and failure management.
  • Clark Lubbers, Fellow – With 35 years and 56 patents pending or granted, Lubbers was an early and influential force in virtualization, performance and RAID architectures.
  • Bill Pagano, Fellow – With 35 years of experience, Pagano was the lead hardware designer of the company’s Intelligent Storage Element (ISE™) technology and holds five patents for his hardware work.

For more information, please visit: http://xiotech.com/press-release.php?id=174.

The Great Shrinking Disc Drive

Feb

14

2010

In the classic 1956 movie, “Invasion of the Body Snatchers” (see the Wikipedia entry below if you’ve never seen or heard of it) the conclusion is left open – for the viewer to decide. That’s part of what makes it so great.

http://en.wikipedia.org/wiki/Invasion_of_the_Body_Snatchers

Now, I can hear you saying…”how the heck is Peglar going to tie that into storage?” Well, work with me here…and let’s look at how the industry is moving. I strongly believe that the 2.5″ disc drive is going to ‘take over the world’ for rotating storage in the not-so-distant future. So, I ask you to look around at your own storage infrastructure, your storage vendor(s) and their roadmap, and see how 2.5″ drives are going to fit into your world…because they are coming, and like the Pod People, you can’t stop them!

The entire idea of using 2.5″ drives in the enterprise has been anathema for many years. Enterprise 2.5″ drives have been around for several years, but vendors never designed arrays to use them, because they not only wouldn’t but couldn’t – they were too small, too fast and too unpredictable. Too small both in terms of physical design (a “disk shelf” wasn’t designed for them) and logical design, i.e. capacity. Too fast in terms of being able to control the physical vibration that ensues, especially for 15K RPM drives, in the drive enclosures they chose to OEM. Too unpredictable in terms of reliability, since most arrays don’t perform any sort of managed reliability – which is critical for 2.5″ drives, even more so than for their 3.5″ predecessors.

Enter the ISE. The ISE was designed with the 2.5″ drives fully in mind. Today, one can have 40 enterprise 2.5″ drives in a 3U container, with full managed reliability and self-healing. The benefits to performance-starved applications are many – but perhaps the best benefit of all is the lack of having to, as a human, deal with herding and managing all these small creatures. This is why no enterprise array vendor except Xiotech has architected the 2.5″ drives into a fully-baked architecture. For others, like the pod people, they are just impossible to control, especially in arbitrated loops.

But the big news of 2.5″ drives is this – the upcoming capacities of these drives are rapidly expanding. Soon, you will see 300, 450 and even 600 GB per drive at 10K RPM, and 15K RPM is being developed as well. Further out, you will see 900 GB and beyond per 2.5″ drive, as bit density improves. These drives, when packaged correctly – as in the ISE – will provide not only a quantum leap in IOPS but a leap in capacity per U – which is a Hobson’s choice today.

I say that because one can certainly package 2TB 3.5″ drives (5400 RPM, SATA, 10^14 BER) in an enclosure, and some vendors are doing just that. However, these drives are extremely difficult to manage in practice – if you think rebuilding a 1TB 7200 drive is fun, you haven’t seen anything yet – and also suffer from poor data integrity, since SATA drives cannot perform DIF and have poor UER compared to their capacity. One colleague at Xiotech, a very senior engineer, correctly said “these drives are a data error waiting to happen.”

The Hobson’s choice is trading off reliability, manageability and human interaction for sheer physical density. It’s a poor choice either way. Having said that, let’s be very clear – 10^15 BER 3.5″ 2TB SATA drives exist, but the storage array vendors aren’t using this technology, going instead for the cheaper route. It’s a net loss for users, truth be told.

This is where the 2.5″ drives really take over. Since they are much smaller physically, the drives/U ratio is superior to the 3.5″ variety, and over time the TB/U ratio will be superior as well. The BER is already superior at 10^16.

So, when you look at what architectures are going to win (take over?) in the future, look at the 2.5″ drives. Like the pod people, they will not be able to be stopped…but the ending is open. The good news is that the ending for us in storage is a happy ending, perhaps unlike the movie. One thing is for certain, though; the disc drive is at once shrinking physically and growing logically, and that’s a movie worth paying for in the theater of storage.

Xiotech and GlassHouse Announce Partnership

Feb

11

2010

Today we announced that we have entered into a new partnership with GlassHouse Technologies, Inc., the leading provider of data center consulting and services. Together, we will engage in a range of joint development and marketing activities designed to deliver scalable, manageable and predictable storage solutions to enterprises worldwide.

“GlassHouse offers Xiotech a proven, established and scalable professional services model for further penetrating the domestic and international markets and meeting the needs of large customers and partners,” said Xiotech President and CEO Alan Atkinson. “We look forward to leveraging GlassHouse’s unrivaled global reach, consulting expertise and umbrella of comprehensive services to quickly serve the demand we’ve seen from enterprise organizations and our partners worldwide.”

For more information, please visit http://xiotech.com/press-release.php?id=173.

Xiotech in 2010 (BD Event Recap)

Feb

02

2010

Last week, several members of the Xiotech executive team discussed the company’s strategy and unveiled a new identity at TheBDEvent in Palo Alto.  Of course, was anyone in the Valley paying attention to anything other than Steve Jobs in the Moscone Center?  In case you were thawed from a cryogenic chamber this week, the iPad is now the coolest must-have gadget in the world and is on my toy wishlist.

Back to work.

Greg and VaNessa Duplessie put on a great show, with A-list content and a storage who’s-who present over the two days.  Of course, cocktail hours and hotel “lobby time” proved equally valuable.  The days were long but stimulating.  I particularly enjoyed meeting some of the industry Twitterati, including @sfoskett, @sunshinemug and @dvellante.  Always good to put a face to a @name, beyond the 70×70 pixel icon.

The Xiotech strategy we outlined is predicated on the fact that we are squarely in the midst of the performance-limited age of storage.  Application owners want more processing power, analytics, online time, transactions…you name it.  And, in order to serve these apps, IT wants to take advantage of innovations like wide-scale virtualization (e.g. for mission critical applications, not just test/dev) and cloud.  Many excellent blogs have been written recently on the subject, including one from David Black, CTO of Oak Investment Partners, and one from Xiotech’s own Rob Peglar.

Also, during the week, the question of how Xiotech will grow was posed to our CEO Alan Atkinson.  He laid out the four pillars of growth during a private luncheon.

“With over 2,000 customers, largely in the mid-market, the enterprise is definitely an important play,” Atkinson said.  Xiotech’s value proposition – price-performance, scalability and ease of integration – is purpose-built for the enterprise.  He also described how desktop virtualization is an imminent phenomenon that will eclipse server virtualization, and how ISE is well-suited to supercharge these projects.  

Second, Atkinson described a significant focus on the channel to serve both U.S. and international markets.  “We grew our channel business nearly 30% in 2009 and intend to increase that in 2010,” he said.  When asked what type of partners Xiotech was interested in, he said that both SI/reseller partners as well as OEMs would be on the radar.

International expansion was detailed as the third pillar of growth.  Xiotech recently signed several new partners to provide reach and distribution.  As Atkinson noted, early signs of growth have already emerged in Western Europe, South Africa, Australia and New Zealand.  “This is particularly important for us as we engage the enterprise,” he said.  “Most banks want to do business with a company who has presence in EMEA.”

The fourth growth pillar was an expansion of the ISE value proposition by cultivating a robust ecosystem.  CTO Steve Sicola detailed the Q1 and Q2 roadmap, including CorteX (coming in Q1) – Xiotech’s RESTful API that will allow developers simple yet powerful access to ISE.  

Oh yeah, in the immortal words of Mr. Jobs, there was one more thing.  In a demonstration of the unparalleled performance of ISE – even in large block parallel streaming workloads like HD video – we announced Mac support for ISE, including boot from SAN and the ability to serve 750 DVD-quality streams from a single ISE without any frame drops.

Jack of all Trades, Master of None?

Jan

31

2010

It’s been quite the ride at Xiotech over the last few weeks.  Today, I have a chance to breathe, think, and blog…in that order :-)

The subject today is very familiar – storage controllers.  The evolution of these controllers has been very interesting, starting back in days of the channel-attached ‘head-of-string’ controllers which included cache for the first time, to better service the mainframe count-key-data storage method.  We had to have caching controllers for decent performance – store count & keys in cache, data down the channel.  It made perfect sense.

Today, we have taken that simple design and subjected it to bloat on an unprecedented scale.  Most of this is due to competitive pressure – vendors feel naturally compelled to engage in feature creep because the competition does so, writing feature D or feature F because they can, keeping up with the Joneses. Then, badly constructed (rigged, some would opine) RFPs merely use these as checkmarks, whether the actual implementation uses the feature or not.  The result is bloat, shelfware and overly complicated software licensing schemes.

I believe the time has come to stop feature creep in controllers, and begin deconstructing the array.  Why?   Overwhelmingly, because the architectural framework of compute and storage have changed drastically over the last ten years.  CPUs, hypervisors, RAM density, bus speed, cost, latency, interconnect, all these factors (and many more) have swung in the user’s favor.  The CPUs in small servers are so fast today that server virtualization is best described as not just mainstream but entrenched, never to be displaced.  Ten years ago, we would have all laughed at the concept of running multiple virtual machines inside a commodity server.  Today, it’s garden variety.  The dream of the mainframe guys has been realized – virtual machines – what a great idea.  It’s 1967 all over again :-)  But I digress.

It’s time to take advantage of the huge cycle count, memory and bandwidth available outside storage controllers.  Let’s return controller design to what it does best, and what it was originally designed to perform – store data, protect data and move data.  All the other functions that a business must perform on it are much more suited today to be done outside a controller.

Consider this.  Why haven’t backup/restore packages – you all know what they are and what they do – been integrated into storage controllers?  Seems like a natural function, doesn’t it?  After all, backup is something nearly every datacenter on the planet does, in some form.  So why do we expect our controllers to squeeze, twist, reorder, shuffle, trade, and perform other seemingly cool/interesting functions that turn out to be ‘features’ in name only – i.e. those designed to increase vendor margin but not increase business efficiency?

The tide is turning.  There are many outstanding pieces of technology now being brought to bear that live outside controllers, are extremely efficient, cost-effective and – most of all – do NOT interfere with store, protect, move.  Let’s use them.

Some would opine that the trick is management of these entities in conjunction with storage controllers (control path and data path) – and I completely agree.  This is why REST is coming to the fore as a management interface for storage.  REST will allow management of all these entities under a single point-of-presence, very efficiently and very openly.  After all, it’s just URIs, HTTP and XML.  Just as Web Services did before it, improving and opening up management, so will REST.

Truthfully, I can’t wait.  I have the luxury of being able to run RESTful interfaces today.  As these are exposed, it’s going to have a great impact on the design of storage infrastructure and datacenter efficiency.

Over the next few years, I anticipate RESTful techniques (such as described in the CDMI spec) to become pervasive, thus opening the door for the storage controller to finally overcome its legacy of being a jack of all trades, master of none.

Xiotech Fuels the “Apple Effect” in New Consumer Media Development, Delivery and Consumption Models

Jan

27

2010

In the wake of Apple’s latest computing revolution, we announced today that our patented Intelligent Storage Element (ISE™) technology has unlocked significant new capabilities in video and audio editing and production, in support of high-compute environments within Macintosh computers from Apple Computer, Inc.

Users are now able to do the following from a single ISE unit:

  • Support the equivalent of operating every movie theater screen in the state of Colorado at the same time from a single storage element.  ISE can drive 750 simultaneous DVD-quality video-on-demand streams without losing a single frame, according to SPC-2 testing.  
  • Play up to 25,000 mp3 files at the same time.
  • Run four professional video editing projects simultaneously from the same console with no frame drops.
  • Move a Mac operating system to ISE and boot from it, which provides a substantial boost in platform performance and reliability.  

For more information, please visit: http://xiotech.com/press-release.php?id=172

2010 so far…

Jan

22

2010

Well, the new year is off to quite a start here at Xiotech.   Year to date: 30,000 miles in the air, major new partnerships underway (watch this space for more details), national sales meeting in Vegas, partner summit in Vegas and Nerdfest in Colorado Springs.  Up next: the BD Event in Palo Alto.  Not a bad first 20 days of the year.

For me, Vegas was the highlight so far.  It was just amazing to see around 400 people from the industry, our salesforce and our partners converge on Las Vegas.  As we unveiled our new logo and Web site, the excitement just built and built.  By the time we left Vegas, the energy was palpable.  It’s great to see the company so fired up, because Xiotech should be fired up.  We have game-changing technology, and people are beginning to understand exactly what ISE means to virtualization, the cloud and other applications that need performance and reliability.  It’s going to be a great year.