NetApp - Standards Watch

Convergence, Cloud, SC09, and the Mainstream

The industry’s premier high performance computing event takes place in Portland Oregon next week in the form of SC09.

SC09 (formerly Supercomputing yy) positions itself as “the international conference for high performance computing, networking, storage and analysis”, and a glance at the conference agenda might convince you that the event is staying close to its roots by showcasing HPC, scientific and technical applications.

However, a closer look also reveals sessions, breakouts and demos on issues and technologies that are now relevant to mainstream IT data centers – from power efficiency, to cloud computing environments, to network convergence and unification.  Though it’s not clear to me that today’s winning technologies for HPC will necessarily be the right thing for the mainstream data center of the future.

A case in point:  the InfiniBand Trade Association will be celebrating 10 years as the leading HPC server cluster interconnect at the show with demos and announcements – see Brian Sparks’ blog on the subject. Despite its success in HPC cluster environments, I don’t believe infiniBand will be a big influence in the mainstream data center of the future. Ethernet (with 10 GigaBit today, and 40/100 GigaBit on the horizon) seems to me to be much better positioned to become the network convergence infrastructure of choice there.

Which brings me to the Ethernet Alliance booth at SC09.  Several member companies will be demonstrating converged storage traffic (including FCoE, iSCSI and NFS) over a 10 GigaBit Ethernet (with DCB) network.  See Jason Blosil’s blog for a good description.

Interesting times we live in …

SNW Europe – Snapshot of the Storage Industry in Europe?

Just recovering from an exhausting few days in Frankfurt at this year’s SNW Europe, so I thought I’d share my impressions while they are still fresh.

 First of all, there seemed to be a real air of optimism around the show. The “quality delegates” (i.e., non-vendor) attendance was up 11% on last year, and as usual the demographic was technically-savvy IT executives.  IT attendees were predominantly German, but there was also significant IT attendance from Europe at large (I talked to IT execs from Germany, the UK, France, the Netherlands, Switzerland. Italy, and Spain).  As usual, there was significant pan-European press and analyst contingent, with 59 press from 13 countries in attendance. 

 The theme of the show managed to retain its predominant focus on Storage while embracing broader Data Center and Virtualization topics and hot new IT trends – for example there was a well-attended Cloud Summit.  There was also a good mix of keynotes from vendors and leading analysts (the Steve Duplessie and Ken Male presentations were particularly good), and educational sessions – as you can see from the agendas for Tuesday and Wednesday.  Also, see here for a listing of the NetApp presenters.

 You can also get a sense of the buzz around the show from the Twitter traffic during the week: #SNWE and #SNWEurope.

 There has been a significant shake-up of the trade media and events business over the past year – as demonstrated by the evolution of Storage Expo into a general IT event (oops – it didn’t evolve in that direction – here’s the correct link!).  Consequently the success of SNW Europe has made it the leading conference for the European Storage industry.  A great show for IT attendees (with high quality presentations), and the leading show for vendors looking to underscore their leadership credentials in the EMEA storage market.

 

SNW Review

On my way back from SNW … so I thought I’d do a quick recap.  Fortunately, much of the work has already been done by the press, analyst and blogger community.

 The big SNIA news was the announcement of the Cloud Storage Initiative, which was covered quite succinctly by Lucas Mearian in ComputerWorld.  The other SNIA news announcements were around SMI-S, Solid State Storage and the recent governing board elections. For details, see the SNIA Newsroom.

 There was a fair amount of product announcement activity at the conference, summarized very well by Dave Simpson in InfoStor.  George Crump gave us an exhaustive rundown of what was going on in his Live from SNW – Day 1 and Day 2 blogs.  In addition, Dave Simpson has a more expansive view of the conference in his blog.

 Of course, if you’re not in IT or the storage industry, SNW might mean something different -- in which case, this is the website for you.

 So, what did I think about SNW? 

 Well, everybody’s first question is about attendance.  My estimate is that IT end user attendance was at about the same level as Spring, which was pretty much level with last Fall.  In terms of vendor attendance, my guess is that it was about the same as Spring, which was significantly down on the Fall show (fewer vendors, each sending fewer representatives).  There were signs of an uptick in vendor activity and engagement, but this was negatively impacted by the schedule conflict with Oracle Open World (which is a major show for many storage vendors). 

 However, judging by the number of product announcements and the jostling for visibility in hot new trends and technologies (cloud, solid state, FCoE, storage efficiency, etc) this continues to be an important event for vendors.

 For IT professionals, it continues to be the best place to get an in-depth understanding of storage technologies, products and trends – judging by the strong attendance in educational sessions, their feedback, and the fact that IT attendance has held up despite the economic downturn.

 Bottom line, SNW continues to be relevant and to provide value to both the IT community and the vendor community. The big question is can it they stay relevant as the industry changes, and more industry interaction is virtual – but that’s a question for all industry events, and SNW seems to be is in much better shape than many.

So, How Big is the iSCSI Market?

My blog last week about the FCoE forecast got me thinking about the iSCSI market, so I asked around a bit about how big people think the iSCSI market is.  Most of them were quite low compared with the numbers reported by analysts.  The problem here seems to be two-fold:  first, iSCSI storage revenues and capacity shipment information doesn’t seem to be very widely reported; and secondly it tends to use infrastructure that is already there. 

 So let’s look first at the latter issue – the infrastructure.  Unlike Fibre Channel, where you can count HBAs and switch ports shipped to get a good idea of the size of the Fibre Channel networking market, with iSCSI that’s much more difficult to figure out.  The majority of people use iSCSI software initiators, which come free with the operating system, and the standard ports or NICS used for general data communications. That means that the number of iSCSI-connected servers is really hard to figure out.  Then at the switch layer, it’s difficult to assess what proportion of Ethernet switch ports shipped  are dedicated to iSCSI traffic.

 This was the focus of a great article by Frank Berry in Byte and Switch last month, “The  Hidden Value of iSCSI Networks”, in which he analyses the situation and assesses the annual networking revenue associated with iSCSI SANs.  The number he comes up with is $585 Million and growing rapidly.  This even surprised me, so I checked with some Ethernet infrastructure vendors who confirmed that the numbers looked about right.  Check out the article and see if you agree.

 Turning to the numbers for iSCSI storage systems, IDC has been tracking this data on a quarterly basis for years now.  Once again, it’s not quite as straightforward as in the case of fibre channel arrays since a significant percentage of iSCSI arrays support multiple protocols, so an assessment has to be made about primary use.  But I suspect that over the years IDC has developed a good methodology to make this assessment.

 In 2007, IDC estimates that 211 PB (211,000 TB) of iSCSI storage was shipped, yielding $789 million in revenue.  2008 saw massive growth with 500 PB (500,000 TB) shipped yielding $1.39 billion in revenue.  The first half of 2009 saw 378PB shipped for $810m – which leads me to conservatively expect an 800 PB (800,000 TB), $1.75 billion year.

 A big market, growing rapidly despite the economy.

Finally, a Realistic FCoE Forecast

A quick follow-up to my mention last week of Seamus Crehan’s analysis and forecast for FCoE deployments … this week I came across an article by Drew Robb who has also seen Crehan’s forecast.  You can find the article here.

Two things strike me about all this. 

The first is that the early analyst FCoE forecasts suffered from the usual shortcomings driven by the new technology hype cycle:

·       Exponential-growth forecast (it’s really easy to do in Excel…)

·       Relatively oblivious to changes in industry economic situation

·       Published before the standard is approved

·       Doesn’t take into account who will deploy the technology or why

… which is probably why Robb starts the article with the observation that “adoption rates haven't yet lived up to the propaganda”.

The second point is that Crehan’s forecast really does seem to take into account today’s realities (requirements, drivers and benefits):

·       Support by all the Fibre Channel infrastructure vendors,

·       Mainstream 10GbE deployment in place.

·       DCB support in “network consolidation” switches

·       Well-documented I/O consolidation savings published by multiple vendors

·       Minimally-disruptive evolutionary path from Fibre Channel network to unified 10GbE network

·       Strong business imperative to pursue IT infrastructure efficiencies through virtualization at the server, network and storage layers

 Suddenly the world seems to be a much saner place…

Busy Week for Ethernet Alliance

Last week was a busy week for the Ethernet Alliance. 

On September 15 they held a one-day “Technology Exploration Forum” to focus on higher-speed Ethernet – Life Beyond IEEE P802.3ba.  By all measures, the event was a great success -  attendance was much higher than predicted, the room was packed all day, and the discussions were lively and engaged.  The high points of the session for me were the end-user presentations:

·       Bikash Kooley (Google) on 100GbE and beyond for data center connectivity;

·       Arien Vijin (Amsterdam Internet Exchange) on 100GbE and beyond from an end user perspective;

·       Mike Bennett (Lawrence Berkeley National Lab) on how higher speed Ethernet could be sensitive to energy efficiency; and  

·       Donn Lee (Facebook) on how their data center could use 100GbE right now – hokey video here.  


(Update) A complete listing of the agenda and the presentations are available here.

And then on September 16, the Ethernet Alliance held their annual members meeting in the Founders Theater at NetApp’s Sunnyvale corporate headquarters.  The high point for this meeting was undoubtedly the excellent presentation by Seamus Crehan of the Dell'Oro Group on the state of Ethernet. 

This is the best analyst presentation I’ve seen in a long time: a thoughtful and rigorous analysis by an extremely knowledgeable person, with detailed Ethernet infrastructure product knowledge, and a thorough understanding of how new technologies enter the market.  His thoughts on 10GbE (now in the mainstream, with perhaps 2million ports expected to be shipped in 2009 -- pulled by on-board 10GBE server ports and the imminence of higher-speed (40/100Gb) standards) and on FCoE deployment were particularly enlightening.  I’m hoping the EA will post his presentation to their website…

Cloud Data Management Interface (CDMI) Spec Released for Public Review

At this week’s Storage Developer Conference SNIA announced the release of the (version 0.8) draft CDMI Spec for public review – just 5 months after the creation of a technical working group to work on it.  The spec is available for review and feedback from: http://www.snia.org/tech_activities/publicreview.

 As I mentioned when I blogged about Cloud Storage standards in April and in June, the pace at which this work is progressing is unprecedented in SNIA, and maybe in the industry. As with any standard, a clear area of applicability, customer-relevant use cases, and broad-based adoption are essential.  The release of the spec covers the first two of these requirements.  The third will depend of relevance to vendor plans. If your company has not been involved, you really should check it out and make sure it works for you.

 For more details, check out the SNIA Insider blog and William Vambenepe’s blog.

 

INCITS Resolves Open Standards Development Controversy

Notable Standards News of 2008 - Part 3 of 3

Last May, I blogged about a controversy raging in INCITS over a rules change on specification access which would have made collaboration with other standards bodies practically impossible by making access only available after the standard had been completed.  The controversy pitted the INCITS Executive Board against the entire membership of the T10 and T11 technical committees, and opened up the specter of SCSI and Fibre Channel standards development looking for a home outside of INCITS.

Fortunately, for all involved a compromise was reached by the end of the year, which makes spec access possible for collaboration during the development process. The new policy and guidelines document is available at
http://www.incits/org/rd1/INCITS_RD1.pdf.

 In my view, credit for this should go to the T10 and T11 members, and particularly to their chairmen, particularly the late Bob Snively, who put in an enormous amount of effort to explain the issues to the EB, seek a compromise, and take on a bunch of extra administrative overhead to ensure the continuation of an effective open standards development process.

Bob gets my ”unsung hero of 2008” award.

IDEMA – trying to avoid a train wreck

Notable Standards News of 2008 - Part 2 of 3

Our industry is often accused of over-hyping emerging technologies well before they are in any way market realities.  Occasionally, however, there is a potentially disruptive change on the horizon that gets no play in the media at all - the news is that there was no news.  This post is about one such development.

Early in this decade 2000, the hard disk drive industry recognized that it would be necessary to make changes to the disk drive data format in order to continue areal density growth while mantainign acceptable reliability.  The Long Data Sector (LDS) Committee was formed under the guidance of IDEMA, and over the next few years they came up with a standard for a proposed transition from 512-byte to 4kB disk sectors.  Details can be found at www.BigSector.org.

Starting in the first calendar quarter of 2011, new drive introductions from disk drive suppliers will ship with 4kB disk sectors.  That may seem like a long time in the future, but this change touches essentially all levels in the storage stack and requires the entire industry food chain to prepare for the transition, from disk vendors, to disk OEMs, bios vendors, driver vendors, OS vendors, tools vendors, and application vendors.

Although all the major drive vendors were involved in the IDEMA 4k Technical Committee, few other representatives of the industry food chain were involved, with the notable exception of Microsoft (who implemented 4kB logical sector support in Vista).

During 2008, IDEMA seemed sufficiently concerned about this apparent lack of engagement by the industry at large that they stepped up their efforts to drive awareness. Mike Fitzpatrick from Fujitsu presented an informative 4kB Data Sector Update on behalf of the IDEMA 4kB Technical Committee at the SNIA Storage Developers Conference in September 2008.  You can find the presentation here.  Check it out.

Let’s hope 2009 sees a bit more focus on the issue…

iSCSI, VMware and Vendor Collaboration

If you haven’t seen this awesome recent post about iSCSI in VMware environments, I strongly encourage you to take a look, particularly if you are running an IT shop and are keen to implement a virtual server environment. 

As the blog title says, it’s a “multivendor post to help our mutual iSCSI customers using VMware”, and I think it’s a great example of vendors co-operating to help the IT community.  Here’s why I think it’s special:

·      It’s a well-written technical best practices piece, with no marketing fluff or spin

·      It’s clearly targeted at typical questions we hear from customers

·      It’s relevant – focused on one of the hottest IT interest areas today – server virtualization, which has been the fastest growing iSCSI environment for some time now

·      The contributors to the post represent a real who’s who of industry experts on the topic:  Chad Sakac (EMC),  Andy Banta (VMware), Vaughn Stewart (NetApp), Eric Schott (Dell/EqualLogic), Adam Carter (HP/Lefthand) and  David Black (EMC).

·      And it’s a real breath of fresh air – in contrast to the constant vendor sniping we see in the industry today

Kudos to Chad Sakac from EMC for reaching out to the other vendors to make it happen!