Data Center Strategies

SMI-S: Missing in Action

At Storage Networking World last week, something seemed out-of-place, but I couldn't quite put my finger on it until I looked at the schedule of events -- then it hit me. The Storage Management Initiative (SMI) was no where to be found. Not on the conference agenda break out sessions, not in the pre-conference tutorials, not at any vendor booth, no where. Basically SMI-S was "missing in action". With all the hoopla and trumpeting of technologies and initiatives like Green Storage Initiative (GSI) and Fibre Channel over Ethernet (See Nik's post on this) the lack of SMI-S coverage would have been easy to miss. This seemed strange to me until I realized something else. The marketing effort for SMI has really dropped off also. This is very concerning for a couple of reasons. First, the lack of marketing means that the value proposition for SMI-S (interoperability, vendor choice) is not getting out. Second, it seems to suggest that the companies participating in the initiative -- I'm talking about IBM, Sun, Dell, Symantec, HP, and others -- are not ponying up the marketing people to champion the effort. Could it be that the marketing people at these companies realize that the value proposition for SMI-S is less than stellar? Or do they realize that working on SMI-S is a low priority effort and amounts to a career cul-de-sac job? Whatever the reason, the conspicuously missing marketing folks for the technology raise a lot of questions and concerns over the viability of the management technology. Without a concerted marketing effort, SMI-S amounts to a technology-for-technology sake initiative.

To be fair, Nik and I talked with Vincent Franceschini (SNIA Chairman) and Wayne Adams (SNIA Treasurer and Chairman Emeritus) about the situation. And to their credit, they both acknowledged the problem and defended the position to not put SMI-S front and center this year at SNW. According to both, the marketing situation is serious and will be addressed within SNIA in the next few months (although no direct plans were disclosed). The decision not to make SMI-S front and center at SNW is a good one I think. SNIA needs to focus on issues that have a high impact on IT industry today; these include lowering IT costs through energy efficient storage, lower-cost infrastructures (Ethernet storage like iSCSI and FCoE) and storage security (Storage Security Industry Forum). And, there is a feeling that SMI-S needs to be moved to the mainstream -- in other words, SMI-S just comes with the product you would buy. In fact, certified SMI-S certifications and implementations are out there (believe it or not). Products like Tek-Tools Storage Profiler are now using SMI-S, and the SMI-S Conformance Testing Program includes several storage products that claim SMI-S certification. But, one-click down reveals more than meets the eye. Notice on the Conformance Testing Program Page that only seven -- yes seven -- clients are SMI-S v1.1 certified. This means that there are only seven client applications that -- in some way -- use SMI-S (and may not meet your needs). Also, the dirty little secret to the SMI-S conformance testing program is that both the client management app and storage product don’t have to enable any real manageability to pass. The client simply needs to be able to discover an SMI-S implementation (CIMOM and provider for a storage product) and read one object. The storage products simply need to be able to be discovered and have one object to read. No real management functionality required. This is a huge hole in the program and probably why so many storage products have the certification.

Now, I'm not one to go off the deep end and say that SMI-S is the worst thing ever, like other analysts might be inclined to do. But, at some point you have to ask, how much steam does SMI-S have left in it? If none of the marketing folks are willing to stick to it, SNIA is ready to move on, and the certification program can allow vendors to present SMI-S as a checkbox rather than full-on functionality, what does that say?

For the customer, all of this may be a no-op. We all know that storage management is an issue, but if a storage management vendor can create a solution can abstract away the intricacies of a HW vendor's unique APIs and provide some level of storage vendor choice then, who cares about the underlying plumbing? And in the end, that may be what keeps SMI-S from making it. But until we have a management application like that in our hands, the jury is still out on SMI-S. And customers, don't be fooled by management apps and storage products claiming SMI-S certification. Any vendor touting SMI-S certification should be scrutinized for functionality. In fact, I'm considering writing a sample RFI for customers to root-out SMI-S functionality claimed by vendors. Would this be of help?

What do you think? Will SMI-S become the storage management API of the future? I'd like to invite members of SNIA and storage vendors alike (especially marketing folks) to chime in. Until then, I'll continue to beat this drum; seeking validation through blogs and our Catalyst conferences.

[posted by: Drue Reeves]

Storage - The Next Hot Target

Last week at Storage Networking World, SNIA announced it's intention to get in the "green" game with the intent of promoting more energy efficient storage technologies with its "Green Storage Initiative". Given the hype around the green data center it's no surprise that SNIA would want to join in, the question is whether they can do much to help. Before looking at SNIA's role it's worth examining the current state of play.

Reducing the power and cooling requirements in data centers has been a major initiative in many IT organizations in recent times. While some have marketed the effort as part presenting a more "eco-friendly" face to their customers, real motivation for this effort comes from two primary considerations:

• Substantial reductions in the amount of energy consumed by IT can have a significant impact on overall data center costs. 
• educing the energy footprint of the existing IT infrastructure may allow additional infrastructure to be added without requiring a new data center.

But regardless of the motives, it's been an important part of data center strategy.

So far most of the reduction in overall energy footprint for IT has come from the application servers. Application servers were chosen because they represented the largest consumers of energy and because they were used very inefficiently. But a side effect of making the server infrastructure more efficient is that other consumers of power now represent a more attractive target for improvement. Next on the list is storage, according to DELL's CTO Kevin Kettler, storage accounted for 37% of the energy consumed in DELL's data centers, vs. 40% for application servers. With application servers becoming more efficient, it's only a matter of time before storage becomes the number #1 consumer of power in the data center.

The industry has attacked the energy footprint for application servers with two basic strategies:

• More efficient hardware - processors, power supplies and other components of the typical application server have all improved considerably. 
• Better utilization - with ever more powerful servers, the strategy of running a single application on each host is extremely inefficient. Server virtualization has enabled huge gains from consolidation.

The gains from hardware improvements are important, but a drop in the bucket compared to the gains from server virtualization. It's likely that improving utilization will also be the most effective approach for storage, there's only so much you can do to reduce the power requirements of a disk! But storage utilization means different things to different people, so we are likely to see a number of approaches to improving utilization, with technologies such as thin provisioning, automated tiered storage, data deduplication, massive arrays of idle disk (MAID) all having a role to play. But there is no one size fits all solution, IT consumers will need to understand where these technology make sense and how they can be combined into a holistic approach to reducing the energy footprint of storage.

SNIA believes educating IT consumers about these options will be an area where it can help, but I'm not so sure! One of the problems with SNIA is that it is  dominated by large vendors who pay most of its bills and have enough staff to dedicate resources to its working groups, committees and educational efforts.

As a result almost everything SNIA does reflects the wishes of its largest members. Consider the "storage virtualization tutorial" introduced at SNW in 2001. The goal of the tutorial was to educated IT consumers about emerging storage virtualization technologies. At the time there were several relatively low cost software/appliance-based systems that were available. But these approaches represented a serious threat to the major storage companies who preferred expensive switch-based solutions. So it's no surprise that the tutorial gave equal time to both approaches, even though the switch-based solutions were half a decade away from a 1.0 release. The tutorial was also unable to make recommendations, or point out potential drawbacks with the different approaches. So the end result was a watered down overview with no conclusions or actionable recommendations for IT consumers.

I fear that similar problems will impeded any effort by SNIA to educate IT consumers about how to reduce the energy footprint of storage. The vendors who control SNIA will shape the message to best fit whatever technologies they favor, and consumers will be left to work out which ones are available and what the strengths and weaknesses are.

Posted by: Nik Simpson

Hang together or hang separately...

While signing the Declaration of Independence in 1776, Benjamin Franklin made the memorable remark "We must, indeed, all hang together, or most assuredly we shall all hang separately." I couldn't help thinking of this remark as I witnessed the spectacle of all the FC switch and HBA players singing from the same hymn book as they touted the benefits of Fibre Channel over Ethernet (FCoE) at this week's Storage Networking World. Such unity of purpose is truly remarkable for an industry where lack of cooperation and cut throat competition has been the rule. But when you look at the situation faced by these vendors it's not so hard to understand.

Put simply, the growing acceptance of iSCSI, combined with the inevitable commoditization of 10GigE represents the most dangerous threat that the Fibre Channel industry has ever faced. If iSCSI running over 10GigE is fast enough for most applications, then many existing customers will skip future generations of FC technology in favor of low cost iSCSI-based storage networks. Worse still, the majority of the market that has never bought FC, never will!  If that nightmare comes to pass, then it would be a disaster of biblical proportions for the entire FC industry.

The beauty of iSCSI is it's ability to run over "lossy" but low cost Ethernet infrastructure because of the ability of iSCSI to easily recover from transmission errors. In contrast, FC requires an expensive and lossless networking infrastructure built around FC switches and host bus adapters.  Fibre Channel over Ethernet (FCoE) is an effort to redesign Ethernet switching to behave more like an FC switched fabric. This "enhancement" would enable the Ethernet network to run the FC protocol directly by simply converting the FC frame to an Ethernet frame inside the switch, thus removing the need for iSCSI.  At first glance this "Data center class Ethernet" as the FC vendors like to call it, seems like a good idea because it:

• Enables storage and data traffic over the same network fabric 
• Preserves existing investment in FC infrastructure and FC SAN management expertise 
• Eliminates expensive iSCSI gateways between FC storage and Ethernet attached hosts

But the vendors are only telling half of the story and it's the half that looks good! What they neglected to mention included:

• FCoE will require expensive lossless Ethernet switching, so the single fabric benefit is largely an illusion. The reality is that most environments would end up with a two-tier Ethernet infrastructure with storage traffic going through the FCoE switches and IP network traffic going through a conventional low cost Ethernet infrastructure. 
• Preservation of existing investment/expertise is only of interest to those organizations that have bought into FC for SANs. It does nothing for all the companies that have shied away from FC SANs because of the cost and management issues associated with them. 
• The iSCSI gateway argument is straw man argument that neglects the fact that the vast majority of modern storage arrays offer native iSCSI connections, which eliminates the need for any gateway in the first place. 
• None of the major server vendors have expressed any interest in making FCoE capable network adapters a part of their future server platforms. In contrast, they all have plans to replace standard 1GigE with 10GigE as soon as the price and power consumption makes it attractive, and those interfaces will have some degree of hardware acceleration for TCP/IP built-in. 
• Proponents of FCoE will undoubtedly tout blade servers as their platform of choice, and some IHVs may create an FCoE I/O module (though this remains to be seen), in addition to a standard 10GigE I/O module . The question will be whether customers will want to pay the extra $$ for the FCoE module in addition to the 10GigE module which will be included in the price?

One thing that can almost be guaranteed is that some existing FC customers will adopt FCoE at least in the short term, and when they do, the FC vendors will trumpet their success and try to position FCoE as a better technology than iSCSI. But any FCoE success will come at the expense of existing FC infrastructure sales and will do little to impede the growth of iSCSI SANs.

So at best, FCoE will appeal to some existing SAN customers as an alternative to FC. At worst FCoE will be seen seen for what it is; a transparent and self-serving attempt by the FC vendors to preserve a very profitable business model.

Posted by: Nik Simpson

A Sign of the Times

I'm going to go out on a limb here and say that Sun has finally done something smart with respect to their storage strategy. For a long time, Sun's storage strategy has been a soap opera that has keep the entire industry entertained. We've had a succession of new leaders, new strategies, and new acquisitions that have have disappeared without trace leaving customers to pick up the pieces and struggle with orphaned hardware platforms. It's easy to forget how profligate Sun has been with acquisitions; remember Encore (1997), MaxStrat(1999), Highground (2000), Pyrus (2002), StorageTek (2006)? All but StorageTek are long gone, the same goes for senior executives in Sun's storage group who come and go like the tide.

So Sun doing something sensible in storage is worthy of comment (even if it is widely believed to be one of the signs of the Apocalypse). In case you missed it, what Sun did this week was to fold it's storage group back into the systems business along with its servers and the Solaris operating system. Big deal I hear you say, after all weren't we due another storage strategy from Sun, it's been months since the last one? But making storage a part of same R&D center as Sun's x86/x64 servers and operating systems makes a lot of sense when you consider:

• Commodity components are an increasingly common part of modern storage subsystems. That means they have a great deal in common with a standard x86/x64 based server platform, which funnily enough Sun already builds. 
• The days of custom, hand-coded, storage operating systems (the specialized OS that manages the subsystem's hardware and services) are over. Specialized code has largely been replaced by commodity operating systems like Windows and LINUX (albeit in a cut down form). Again, Sun has a pretty good OS to run on it's next generation storage platforms in the form of Solaris x86. 
• The main area for future growth in storage sales will be to the SMB, where there is considerable scope for innovation in areas of storage management and services. The best platform to pursue that market will be one built on commodity components and software, not custom hardware. 
• Building server and storage platforms on a common x86/x64 hardware platform will help reduce production costs for the platform by increasing volume. 
• It gives Sun the opportunity to improve the integration between server and storage management. That's something all the server and companies need to do.

So absorbing the storage group into the systems business makes a lot of sense in the context of how the next generation storage platforms will be built and the markets they will address. Whether Sun can capitalize on the synergies between their server, system software, and storage business is open to question, and to expect immediate success would be, as Oscar Wilde said, "a triumph of hope over experience."

The interesting question in all this is what happens to Sun's latest storage acquisition - StorageTek (STK). STK bought a number of things to Sun, most notably a pretty health tape-based backup business. This commodity play works well for that backup market, where an increasing part of the backup load is handled by commodity server platforms running things like Virtual Tape Library or Disk-to-Disk backup applications.

The other significant piece of STK was a midrange array business built around controller technology OEMed from LSI. This is a platform that is ripe for replacement with a commodity hardware/software platform. So the big question is will Sun replace conventional arrays with a commodity hardware/software platform or will they continue to make traditional arrays for the midrange market?

I'm not brave enough to try and predict where Sun will go with midrange arrays. remember this is a soap opera, and unexpected twists, cliffhangers and unlikely coincidences are all part of the fun!

Posted by: Nik Simpson