Data Center Madness

It’s been an interesting first quarter. Cisco and HP dumping on each other last month, 3PAR turns up the heat in automated storage tiering and then HP’s two-day analyst event in March, followed up by EMC’s grand vision for federated storage and then LSI’s analyst event in NYC. Next week is Iron Mountain in Boston where we’ll hear about the Mimosa acquisition. Then it’s SNW in April which will bring a ton of announcements. 

It looks like 2010 is shaping up as a good rebound year. Earnings reports in December were optimistic and the year-on-year comparisons with Q1 2009 should be excellent because Q1 2009 was so miserable. Tech companies are exiting the downturn with outstanding balance sheets. Keep an eye on the relationship between housing starts and unemployment. 

Major themes have of course been virtualization and cloud computing. The Wikibon community heard in early March from business continuity practitioners how the cloud has the potential to transform disaster recovery and business continuance, dovetailing with EMC’s vision of federated storage. 

Shifting Data Center Landscape 

The subtext to all this high level action these past couple of months is a shifting landscape in the data center. We’re seeing big players like Oracle, IBM, HP, Microsoft, Intel and Cisco position for the next wave of innovation; and we’re seeing companies like VMware and EMC trying to get a seat at the big dance…while NetApp stays focused on doing what it does best—making storage simpler. 

The sub-sub-text here is the competitive developments in the Fibre Channel business. The vendor community wants to keep users locked into FC without it looking like a lock-in so they’re giving customers a dual FC and FCoE path. Users are uneasy about FCoE but ultimately they will have little choice to adopt as costs come down and the technology moves from CNAs in servers to top-of-rack switches and ultimately converged switches and native FCoE devices. While networks might not converge in the near term, technologies will—around 10 Gb Ethernet and FCoE. 

Fibre Channel Ripple Effects 

We’re also seeing a new wave of competition in the FC switch business. HP recently dumped Cisco (they’d never admit this) and began private labeling QLogic FC switches. Yesterday QLogic announced that its switches are seeping into EMC (EMC is of course playing Switzerland saying it sells Cicso, Brocade and QLogic products). HP’s move to private label QLogic switches essentially funds QLogic’s entry into a market that’s dominated by Brocade (~80% market share according to Dell’Oro). Every market needs competition and the FC switch business really needs more competition. 

Of course QLogic is only focused on edge switching and doesn’t have an edge-to-core offering so perhaps the threat to Brocade is benign. But something’s happening in this space and it’s important because a lot of these big data center, federated, cache coherent installations will be running on FC or FCoE. FC won’t die. Wikibon estimates there are $25B in FC assets installed and CIO’s are not going to rip and replace that infrastructure—it would be crazy. That means there’s a good business still to be had and because it’s a mature business the logical way to grow is to gain share. Who has the share? Brocade. 

Wither Brocade?  

Two blogs got my attention yesterday. First was one posted on Silicon Angle pointing out that Brocade is under pressure. John Furrier wrote: 

Recently Brocade has been under fire for lack of performance due to pressure from Juniper and QLogic.  Vendors are broadening the ecosystem to create more competition and accelerate the convergence of storage and networking and servers. 

I doubt QLogic has had an impact on Brocade yet but there’s clearly some friction brewing there. But in another blog yesterday, StorageMojo’s Robin Harris was even more cutting. Entitled Brocade’s unraveling, Harris argued that Brocade’s strategy has painted it in a corner because its OEMs had diverging objectives to those of Brocade. Harris wrote: 

Brocade’s troubles reflect the dangers of an OEM strategy when your partner’s strategic interests are different than yours. None of them wanted Brocade to succeed as a networking company. 

The FC business is not growing. FCoE is a measure to keep the business alive for another 10 years and lock-in customers to a path that is safer than alternative protocols. Sure, eventually the world will go all Ethernet—the operative word there is eventually. The only ways to grow in this business are to raise prices or steal share from Brocade. And because Cisco is at war with the world it seems as though companies like HP are willing to become benefactors to QLogic which creates more pressure on Brocade. The company is getting squeezed by a backlash from SAN complexity and competition from Cisco, QLogic and Ethernet generally. What’s more, no one seems to want to shell out the $2.5B+ to buy Brocade even though the company has been shopping itself. Yes it has cash on the balance sheet (~$500M) but its debt is also high (~$1B). 

Complexity 

All this near term posturing underpins the bigger issue—users want things to be simpler. SAN was supposed to deliver that. It didn’t. And now there’s so much installed storage network infrastructure that users can’t get rid of stuff easily. Everyone understands that the ‘Google Effect’ is changing everything and all companies need to figure out how to simplify infrastructure management. That’s why all the arms dealers are trying to figure out an Ethernet strategy; EMC is riding VMware and guys like HP are trying to add value by simplifying outside of the hypervisor. 

Bottom line: Simplify or die…what’s your strategy?

LSI in the Big Apple

Why does LSI exist? In 2007, Bill Zeitler, former head of IBM’s Systems and Technology business said something to me that summed it up perfectly: “Industry revenue is growing at 9% and R&D spending is growing at 12%– we can’t afford to fund everything.”Bill retired a couple of years ago and the numbers have obviously changed—but the pressure to focus R&D dollars hasn’t.

So you want to be in the OEM business and help guys like IBM out? Why not? A deal with an IBM or HP can mean hundreds of millions of dollars in revenue and to a smaller company it can provide an enormous boost in valuation. The mindset of some companies that want to sell through OEM channels is “hey – let’s hire an OEM sales rep…someone with lots of contacts.” This mentality says we have a technology. It’s a hammer. Every OEM opportunity is a nail.

From what I can tell, this is not how LSI does business.

LSI is a company with $2.2B in revenue, gross margin in the mid-40’s, nearly $1B in cash and 5,400 employees. LSI does hard stuff like 40nm Physical Layer (PHY) macrocell technology, storage and networking silicon, other custom silicon, disk drive read channels, virtualization software and a boatload of other technology innovations that are difficult to build. It doesn’t sell direct.

Its business model is to invest in building technologies for OEM’s and then essentially syndicate its R&D expenditure across a large number of customers. Its customers get the benefit of lower R&D and they can still fill holes in their product lines. LSI’s willingness to customize is how its OEM’s can differentiate (if they can justify paying for that innovation). They also organize to service the heck out of OEM’s.

The Highlights

Here’s my quick take on the event and its content and a short video with Steve Fingerhut of LSI:

• Grand Hyatt. NYC. Very good food. Crappy Internet. First class event. ~60 analysts by my count. Building nicely from “starter event” two years ago at SNW and last year in Allentown.
• Major themes – financial turnaround; restructured the business. Growing again, nicely thanks to an OEM recovery in Q’s 2-4 in 2009. Strong outlook for 2010.
• New 6 gig SAS switch to enable shareable DAS. Target is for small-to-mid sized businesses that can’t deal with the complexity of SAN. Good idea. Storage function has been migrating from the array back to the host—being pulled along by Microsoft and Oracle. Storage vendors need a strategy to deal with this.
• Big SAS and shared DAS message (e.g. SAS switching) – Playing off of Microsoft and Oracle themes to aggressively commoditize SAN-based storage. LSI plays both sides of that fence with an increasingly strong SAN offering (e.g. SVM and other products).
• Lots of emphasis on SSD, including a new PCIe SSD–similar to Fusion-io with a demo showing higher performance than Fusion-io. Despite the demo, LSI is not shipping yet and Fusion-io has a bunch of real customers and has been shipping for years (2-3).  Also, my bet is Fusion-io will have a new version of its solution out at some point that will leapfrog the LSI technology. Observers who watch this space will tell you correctly that raw performance benchmarks don’t usually tell the whole story because of things like context switching or “IO storms.” Users need to take into consideration the full application picture.
• Big emphasis on scale out file and unified file and block; enabled by ONStor acquisition. Scale out NAS is a big theme this year with Ibrix/HP, IBM’s SONAS, Isilon gaining traction and NetApp perfecting ONTAP 8. LSI’s strategy is, as one would expect, to focus on the midrange and leave the big honking NAS to guys like IBM.
• Another key message was LSI’s ability to integrate its own as well as third party IP. It uses an abstraction layer in its architecture to make IP more portable.

What was Missing?

It’s not all perfect at LSI. The company is often late to market with array-based innovations. Thin provisioning and storage virtualization are an example. The company has no data reduction (e.g. deduplication and compression) story and I thought the emphasis on VMware was light; other than a vCenter plug in demo. I would have like to have heard more about VMware API integration, especially around VAAI and VADP.

I’m hearing from some suppliers that they still don’t have the most up-to-date SDK’s from VMware -hmmm…wonder if the folks at Hopkinton have those yet J. This could be handcuffing LSI to an extent as I’m sure from VMware’s perspective, LSI is in line behind EMC, NetApp, HP and IBM.

I recognize that LSI is pretty good with integration and integration takes time. Many companies acquire firms and just keep selling whereas LSI really emphasizes integration. Nonetheless, with the re-structuring of LSI’s consumer and mobile business behind it, I’ll be looking for the company to be executing more as a market leader rather than following the trends.

I also spoke with CFO Bryon Look about the company’s gross margins which hover around 47%. For all the hard stuff LSI does you’d think margins could go higher. Of course Bryon couldn’t give me what I wanted but he did say that as the product mix shifts to software content LSI should be able to maintain its targets—which are in the mid 40’s. From my perspective if Emulex and QLogic can get 65% gross margins for the difficult stuff they do then LSI should be able to crack 50 points. But selling to large OEM’s is never easy.

On balance a very good use of my time. I think LSI’s model is coming together nicely and its business approach is by far the best OEM reference model in the storage business.

Wow - HP Finally Gets It

HP is finally getting its act together in storage. For years HP has talked about leveraging the systems and storage business but that vision has never materialized as a serious differentiator. Guess what? It’s finally happening; And there’s an added bonus, HP’s move to compete with Cisco makes it the only company on the planet that owns a robust server, storage and networking stack.

How did that happen? Here’s the simple formula:

• Hire someone (Donatelli) that understands the business and can execute
• Keep the best executives, restructure the talent pool and bring in people you trust
• Create a vision with big time goals
• Set some specific, measurable objectives and put a strategy in place to meet these
• Hire more great people – including a boatload of feet on the street
• Organize and incentivize them properly
• Acquire Ibrix and 3Com.
• Execute

That’s the play.

Having just spent two days at an analyst event I came away much more enthusiastic about HP than I have in more than a decade. I witnessed a mashup of matured EMC testosterone with HP professionalism and management excellence. Throw in an enormous dose of $50B supply chain muscle and you have a serious player with a new kid on the block mentality.

HP Goes Commodity

Someone said to me the other day (trying to disinfect my recently Donatelli-ized brain) “HP is going commodity – they’re hosed.” My response: When Mark Hurd took over HP its PC business was sucking wind. Last quarter it threw off about $500M in operating profit. If that’s commodity, I’ll take it. Shades of Jerome B. York people—cut the fat; print cash.

Everyone on the planet is talking about modular components. But from what I saw this week HP is way ahead in terms of developing a single modular platform that combines servers, storage and networking. EMC’s doing some good work between midrange and V-Max. Cisco Nexus – nice package. IBM blades…good stuff. But HP is combining server, storage and network packaging in a single platform. And with the largest supply chain leverage in the world it will do some damage. Son of Storageworks x 3 folks.

HP Going for Growth—Low end Block and Scale out NAS

Donatelli and Roberson both cut their teeth in a high end block-based world and they have their sights set on growth– Ironically, low end block with Lefthand and scale out file storage with Ibrix. The product strategy is to have a single modular platform with common components; make the storage controller a commodity server and give the system personality through software. Block, file, name your protocol, scale out, single name space, virtualize the IO and network, differentiate with flexibility and responsiveness.  Oh and did I mention cost structure is important? Shades of NetApp with a $50B supply chain leverage.

HP’s version of V-Block with a $50B supply chain. That’s equal to the combined revenue of VMware + Cisco + EMC. What about VMware integration you ask? Fair question– how does HP differentiate?

What About Differentiation?

Over lunch at the analyst meeting I was speaking with Christine Reischl, HP’s Senior Vice President and General Manager of Industry Standard Servers (ISS). She flat out stated that for HP, virtualization has to be about more than the hypervisor; that HP needs to find ways to differentiate through packaging, cross server/storage/networking group synergies, virtual IO—i.e. unifying management and IO, scale out, mission criticality, energy efficiency, etc.

Here’s another form of innovation. HP has struck a deal with the Microsoft devil to make purpose built appliances for Microsoft Exchange, SharePoint and SQL. Microsoft’s head is up its DAS but HP doesn’t care because it has a deal to sell Exadata-like appliances along with Mr. Softy.

And once HP completes the acquisition of 3Com it will be a very solid #2 in networking in the data center behind Cisco. Can HP compete with trying-to-re-invent-the-Internet Cisco? You bet. While Cisco is announcing that the CRS-3 will change our lives, HP is plotting to cut data center networking costs and simplify the lives of IT managers across the planet. Is there a market for that? Um. Ya.

Any Gaps?

Of course. HP’s storage portfolio is still fragmented and needs to be rationalized. It has to transition from a block-based EVA installed base to a new modular unified storage world. HP appears to have missed the data deduplication boat and needs to figure out the whole data reduction thing. Of bigger importance, it doesn’t own the hypervisor and is out of the running (for now) in the really big software land grabs. Software represents only about 3% of HP’s revenue yet the company continually touts taht it is the 6th largest software company in the world (this is based on U.S. revenues by the way)– HP can do better in software.  And the EDS integration is still taking a long time; although EDS is a huge asset long term, especially in vertical industries for HP.

But the bottom line for me is the new organization. Dave Donatelli, Randy Seidl, Garry Veale, Tom Joyce—all ex-EMC, all really competent, well connected, motivated and serious players. HP’s ESSN (Enterprise Storage Servers and Networking) is hiring hundreds of sales people and they report to Donatelli. Inside of twelve months, Donatelli has made this happen plus acquired Ibrix, is in the process of acquiring 3COM and has struck an important deal with Microsoft.

I can’t wait to see what happens in the next twelve months.

Exploting Distance as a Business Capability

Last October I wrote:

By late 2012, federated storage will be the architecture of choice for large new storage deployments. These capabilities will dramatically improve IT’s ability to respond to business needs with minimal disruptions. IT organizations should plan to aggressively adopt federated storage as it becomes commercially available.

Around that same timeframe, Nick Allen wrote a piece about Chad Sakac’s Long Distance Live vMotion demo at VMworld 2009 and asked :

So, has EMC solved the long-distance cache coherency and distributed lock management problem that has plagued the industry forever?

Today (3/11/2010), EMC’s President of Information Infrastructure, Pat Gelsinger unveiled a vision for federated storage which makes our two posts last year pretty timely. You really have to hand it to EMC. In late 2007 things were getting kind of boring and Joe Tucci leaks Hulk and Maui. Then a few months later, EMC lands a haymaker by announcing flash drives in Symmetrix. Next was the buzz EMC created around its private cloud vision; followed by Fully Automated Storage Tiering (FAST).  And now– virtualized, private cloud-ready, dynamically scalable, distributed cache coherent, automated, always-on, efficient, on-demand, active/active, globally federated storage. (Thanks to Chuck Hollis for providing Gelsinger’s slides which enabled that sentence).

What is Federated Storage?

Here’s Wikibon’s definition:

Federated storage is the collection of autonomous storage resources governed by a common management system that provides rules about how data is stored, managed, and migrated throughout the storage network. In this definition, storage resources include disk capacity managed by controllers or appliances controlling multiple arrays.

Practitioners should think of individual resources (e.g. arrays) as nodes within the federation.

Shades of Global Sysplex

Global Sysplex was an IBM mainframe innovation with the following attributes:

• A geographically dispersed set of processors within synchronous distances.
• Separate resources acting as a single global pool.
• Applications and data could be moved around and shared across processors.
• One key ingredient of Sysplex was the Sysplex timing mechanism which acted as a time stamp (not as important for federated storage).
• Another element was global cache management to store things in a global shared cache (important for federated storage).

By enabling a loosely coupled set of storage resource nodes to act unilaterally and still be managed centrally, organizations can create networks of virtually limitless capacities, move data and applications globally, eliminate disruptive migrations and dramatically improve recovery.

Oh…and this time you don’t need a bazillion dollar mainframe complex to do it.

What Problems Does this Solve?

Today’s remote replication systems (e.g. SRDF) can be incredibly complex and expensive. Organizations need to set up disk systems at multiple locations and pre-allocate space and communications lines and essentially purpose build replica systems to support specific applications. Federated storage enables a general purpose virtualized pool that can be established to migrate workloads, free up disk space, end of life assets and bring on new resources much more quickly and at substantially lower costs. Planning is simplified and RPO is improved.

The big benefits will be at synchronous distances but federated storage will also help improve the efficiencies at asynchronous distance. I wouldn’t hold your breath waiting for asynch active/active (remember Chad’s demo took a long time to move a few GBs – how long to move a TB?) but the overheads of wait times and acknowledging writes will be dramatically improved by avoiding time-consuming remote gets and message queues.

And the services delivered can be way more leveraged across the application portfolio.

Questions and Comments from The Blogosphere

How does this all work?

Gelsinger mentioned that one of the secret ingredients was IP acquired from Yotta Yotta. @RayLucchesi wrote on his blog today:

Yotta Yotta’s product was a distributed, coherent caching appliance that had FC front end ports, an Infiniband appliance internal network and both FC and WAN backend links.

My understanding is this isn’t just Yotta Yotta re-purposed but that IP was the starting point.

Ray also said:

So it looks like #emcvs is solving the DaaD problem. makes active-active and vmotion ovr dist easier but is their a mkt for it?

I think so. First, most organizations in the F500 doing SRDF will look at this over time and see it as a more attractive alternative. But I think this vision has broader appeal than just the F500 because it can potentially scale down, over time.

The best comment of the day on Twitter goes to @storageio (Greg Schultz).

Given Yotta Yottas history of delivering Nada Nada, can EMC turn it around with #emcvs into Lotta Lotta?

I have no idea but that’s pretty funny.

The Top 10 playing in the Innovation Sandbox were:

RSA 2010 Security Innovation Sandbox

Five of the ten finalists are focused on the complex issues associated with securing cloud-based services and applications.  HyTrust and Catbird Systems have similar but different offerings in VM security policy automation, with HyTrust focus on centralizing the automation of access control, patch management, separation of duties across infrastructure components.  Catbird consumes attributes associated with the security of a VM environment and maps these attributes to standard control and audit frameworks.  Navajo Systems is offering a security appliance that users can deploy to encrypt their data transmitted to and stored in SaaS-offered systems.  Navajo claims the ability to index and search encrypted data.    KikuSema is piloting next generation password systems for mobile and desktop applications.  Silver Tail forensic and behavior analyst software is aimed at web-borne and desktop-launched malware by mapping and templating web site navigation and then detecting and alert users of abnormal behavior.

Two important developments to keep track of:

A6, CloudAudit

The goal of CloudAudit is to provide a common interface that allows Cloud providers to automate the Audit, Assertion, Assessment, and Assurance (A6) of their environments and allow authorized consumers of their services to do likewise via an open, extensible and secure API.  CloudAudit is a cross-industry effort of experts in security, audit, assurance and architecture backgrounds.  A6 includes providers and consumers of cloud services and related technology and advisory service firms including Cisco, Sun, Akami, Unisys, Microsoft, VMware, EMC, Rackspace, and over 200 other individual participants, including The Wikibon Project.

The goals of the A6 Cloud Audit Group are to:

• Utilize a security and audit automation capability which existing tools/protocols/frameworks via a standard, open, and extensible set of interfaces; In sum, a CloudAudit protocol.
• Keep the interfaces lightweight and easy to implement; offer primitive definitions & language structure using HTTP(S)
• Allow for extension and elaboration by providers and choice of trusted assertion validation sources, checklist definitions, etc.
• Not require adoption of other platform-specific APIs
• Provide interfaces to Cloud naming and registry services
• Encourage adoption by driving client side uses; have users “push” providers to opt-in.

ENISA CAM

The Common Assurance Metric (CAM) of the European Network and Information Security Agency is a global initiative that aims to produce objective quantifiable metrics on cloud security and performance, to measure security and auditability of cloud service providers,  the maturity of information security maturity in cloud, third party service providers, and internally hosted systems. This collaborative initiative has received strong support from public and private sectors, industry associations, and global key industry stakeholders.

How is the scope of Audit and Security Different?

There’s a lot that’s been written about cloud security, so how is cloud audit different.  Here’s how.  The role of the audit function is to provide an assessment, independent of other functions like security, of the state of controls in and around an enterprise, and to do testing sufficient to determine if controls are operating effectively and efficiently.  The importance of this role is that it is independent from day-to-day operation.

So what changes from the cloud computing model are most important to auditors?  Here are three:

• There’s less control and direct access to data and applications
• Traditional audit tools (e.g., audit and monitoring software) are not yet integration into cloud services
• The environment is new, and audit techniques, best practices, can control metrics for cloud based services are just starting to develop

The Bottom Line

At the end of the day, the journey to the cloud includes the virtualization and maturity of the audit function.  Why should auditors be the exception?  And the work of the A6 Group and ENISA CAM are positioned to be important vehicles on the auditors journey.

However, the common perception is that energy efficiency and high performance are incompatible objective. That is true with traditional storage architectures but there is an innovative storage option that blends the attributes of high performance storage with energy efficiency.

Atrato recently started shipping their Velocity 1000, a 3U storage module with a unique packaging philosophy that accommodates 160, 2½” drives, creating a spindle density that delivers 16,000 IOP’s. That is an unmatched 5,333 IOP/RU and with a bandwidth approaching 2GB/s. This performance is delivered with an energy requirement 50% to 80% less than other comparable solutions. A recent press release from Atrato Inc describes how one of their customers, The Department of Energy, implemented their solution to help solve both a green data center objection and to eliminate storage IO bottlenecks the were experiencing as they implemented VMware and a growing virtual desktop infrastructure.

Metrics to measure and monitor performance are well known however when thinking through green solutions such metrics are no so well understood. There are two important aspects to consider. First, it is imperative to understand how green will impact the bottom line, and how it will bring value to the data center. Second, it is important to have a working understanding of how to measure green for data storage solutions. These two goals will help IT professionals make intelligent decisions on how to best use green to increase productivity, save money, and reduce management complexity.

Check out this white paper “The Greening of IT” for more detail.

Check

When Dave Donatelli left  EMC last April to run HP’s infrastructure business I wrote at the time that this was another move on the chessboard– “there’s more to this match than storage; and HP just moved a pawn on the board. The question is what can the Rook now see that it couldn’t before?”   

At the time I wrote that the data center business has become an oligopoly where Cisco, HP, IBM, Intel, Microsoft and Oracle are $100B players (revenue and or market cap) and any move they make has an impact that ripples through the industry. I’m putting VMware and EMC in that mix too given the momentum that VMware has.   

You don’t have to squint too hard through the activities to see the impact that integration is having on this industry, particularly the intersection of compute, storage and networking. HP’s initial networking move with ProCurve, and Cisco’s subsequent aggressive entry into servers; along with the VCE announcement has created tremendous friction between two industry giants that came to another head yesterday.   

As far as I can tell, tech blog Silicon Angle broke the story that Cisco was dumping HP as a preferred partner. Cisco then was forced to respond by posting a blog on its web site confirming the story with a video of Keith Goodwin SVP of Cisco’s Partner Program laying the corporate speak on heavily with phrases like:   

“Change is the only constant in the IT industry.”   

As the industry moves toward “collaboration, virtualization and cloud-based services…”   

Cisco wants to “Align with partners who share our network-centric vision”   

“Evolution of our relationship with HP”   

“Top priority is our customers…amidst shifting industry dynamics”   

In the middle of this discussion, Goodwin indicated that Cisco was not renewing its integrator relationship with HP—the largest company in the industry and a huge reseller of Cisco gear. This on the heels of HP late last year, nixing its Cisco Alliance Managers—my understanding is dozens of individuals out of a job whose role was to push Cisco equipment.   

The other shoe to drop today was an OEM relationship between QLogic and HP where HP is private labeling QLogic stackable FC switches (which David Floyer wrote about last month). The word is HP is popping Cisco out as second source to Brocade and slotting in QLogic. QLogic? Yup – QLogic; which is very interesting that this adapter company is now in the middle of this rift. Supposedly Cisco wouldn’t agree to a private label deal with HP so HP said “forget-about-it” and brought in QLogic which is nowhere in the FC switch business and overnight got a ticket to the dance.     

So here’s the deal as I see it. This is a classic case of screw me? Screw you screw me! Two $100B children can’t find room in the IT sandbox so they’re declaring all out war through cliché’s and high fidelity corporate messaging. Internally I can only imagine the invectives that are floating in the hallways of these two companies.   

I’ve seen this go both ways…a company full of hubris shoots itself in the foot with a big partner. I’ve also seen a company back out of an OEM relationship and successfully leverage its channel to compete with the OEM (e.g. EMC/HP).  My guess is in the near term, Cisco will easily make up the couple of hundred million dollars in lost HP business but over time, HP will fine tune its model to compete with Cisco more effectively in networking.   

Meanwhile, HP is playing sugar daddy to QLogic to subsidize its entrance into the FC edge switch market—where there’s hardly any competition today. QLogic is a non-threat to HP and will happily do whatever its new benefactor demands as long as it helps the company grow.   

As the HP Cisco relationship unravels the race is on for Cisco to integrate its stack and maintain leverage while gaining a foothold in servers; while HP tries to integrate its compute, storage and network assets and keep the aggressive Cisco out of its server installed base. 

While these two giants battle it out, keep an eye on Chessmaster Ellison as overnight, Oracle has built one of the most robust stacks in the business and has huge disruptive potential.

Which Way to Turn?

Storage function is slowly and steadily migrating back toward the CPU supported by flash technology. Application developers are viewing this as a long-term opportunity to deliver substantially higher performance to users and add significant business value. This does not spell the end of storage companies– it’s an opportunity. But it does signal a shift in the technology and business model. and I believe technology suppliers that understand the trend and hedge their long term bets will benefit.  

Background   

For the past fifteen years or more, we’ve seen a steady migration of storage function from the host out to the storage array. It made perfect sense. With several flavors of Unix and NT/Windows-based servers exploding onto the market, the world needed a better way to connect storage to servers and share all the data that was being created. Despite the relative slowness of spinning disk, pushing storage function to the other side of the channel offloaded CPU cycles which were being rapidly consumed in an effort to support emerging client/server applications. It all started with the simple concept of open systems connectivity and compatibility.   

Moving beyond mainframe connectivity, EMC’s Symmetrix was the first system to popularize this trend and its major claim to fame, in addition to performance was its ability to connect to a variety of host systems and support different OS types. EMC engineers worked extremely hard to get to market first with a system that allowed multiple host CPUs to connect to a large shared Symmetrix. The main use case was consolidation and typically the system supported a homogeneous set of servers (e.g. all HP/UX or Sun Solaris). Over time, as SAN became more popular, system connectivity became more heterogeneous, storage function and value increasingly moved outside of the channel and an industry flocked to the opportunity.   

Replication, copy services, data migration, storage management and disaster recovery became function that users were happy to pay for, either directly as a licensed product or as a bundled feature of a storage array that justified higher margins for the vendors. The industry has made billions of dollars in revenue from this movement of function.   

Changes Ahead   

Slowly and steadily we’re starting to see a reverse of this trend in some large and small pockets (see figure); particularly driven by Microsoft and Oracle; but also system vendors like IBM. As well, some startups such as Fusion-io are hoping to take advantage of this movement in the hopes of disrupting current models for application deployment. These vendors, especially Microsoft and Oracle, are eyeing storage function as an opportunity to add value and place more function into their software stacks. Microsoft Exchange, Data Availability Groups (DAG) –which perform Exchange recovery– and Oracle’s Automatic Storage Management (ASM) for 11g –a vertically integrated volume manager– are good examples of this trend generally. Oracle’s Exadata is even more interesting to me because it uses Sun’s on board FlashFire technology (persistent flash on the other side of the channel). Exadata also uses columnar compression to optimize capacity. Exadata is super fast and efficient. It’s also expensive but Oracle is the king of selling on the basis of business value and is also the dark prince of user lock-ins. Oracle will push Exadata as the shining example of the integration of  software and hardware. Ellison has frequently used Apple’s iPod integration as a reference model for better, simpler technology and with Sun in his portfolio now, this is a key strategic initiative and proving ground for the data center. 

The ace in the hole for disk storage has always been that despite its order-of-magnitude slower performance relative to memory speeds, it is persistent. So it can be used for longer term storage and storage networks made it easy to share data stored on disk arrays. The only way to get to that persistent shareable data was to perform an IO. But as the saying goes, the best IO is no IO. With flash, everything changes. We can now place storage function on the CPU side of the channel and allow the application to access a persistent resource at memory speeds. This means that application developers can write code assuming they have access to a single level store without the need to go outside the channel and perform a disk IO– speeding up performance dramatically. Yes it will require a re-write of applications but the business value will be enormous in certain cases (think financial or other database transactions). 

How Real is this Trend?  

I’ve had many discussions with various technologists about this trend and their reaction has ranged from total self-serving agreement to complete self-serving skepticism, to thoughtful discussions about how to take advantage of the trend.  The achilles heel of this scenario is sharing and protecting data that sits in an isolated flash resource? That’s why many have said this technology will support read-only data. But the idea of  providing the notion of “Flash Coherence;” versus cache coherence, across CPU’s in a complex or a data center, through software function, while challenging is interesting nonetheless. I don’t see why write-intensive data are not candidates for this longer term, assuming someone designs a means to protect and share the data in flash. Specifically, I think there is an opportunity for suppliers to become the new Veritas, providing heterogenous services for persistent storage to support an emerging breed of high performance applications? I think it’s worth a discussion. 

What’s also interesting to me is that when I speak to practitioners about infrastructure, they all say they want to simplify. Moving function back to the host is seen as a means of simplification. Microsoft and Oracle clearly support this type of trend as anything they can do to commoditize storage hardware and suck value back to their stacks fits nicely with their agenda. And they have the ear of application heads. I’ve spoken with many who say they understand that placing function in the application stack is more expensive; but they don’t care. They care about the end user experience. 

I was speaking to a Wikibon member recently who is the head of application development at a bank. I asked him what  matters most to him when he’s thinking about designing applications. His answer: “What matters most to me? #1 Availability; #2 Availability; #3 Performance; #4 User experience.” I said what about cost and he said: “Cost? Sure; That’s important to someone but I really don’t think about it much.” 

To Oracle, those are beautiful words and to storage network suppliers who are selling on the basis of shared infrastructure and asset leverage that’s a big “uh oh.” My advice is rather than buck the trend or try and end run the head of A/D and tattle to the CIO (who doesn’t care that much about infrastructure anyway), try and figure out a way to embrace this trend and hedge your bets. 

My colleague David Floyer has written about these trends from a more technical and cost perspective and these additional research pieces may be useful in your thinking about this issue. Please let us know your thoughts on this trend and help us evolve the research for the Wikibon community. 

The Impact of Flash on Future Architectures 

Enteprise Flash Cost Projections