Data-driven

I ended my last blog post with "Integrate and prosper." Little did I know that Cisco, EMC, and VMware were about to unveil a Virtual Computing Environment (VCE) early the following week, the biggest cross-vendor integration project yet seen in the world of computing. Yes there were rumors about a Cisco/EMC joint venture that would sell Cisco servers packaged with EMC storage, but none that I heard captured the boldness and scope of VCE.

The core VCE compute platform is called a "Vblock," an integrated, pre-packaged IT solution consisting of server and networking resources from Cisco; storage, security, and software-based management tools from EMC; and an OS platform (vSphere) from VMware. To market and support Vblocks, the coalition has created two separate entities: a Solution Support Team staffed and funded by the coalition partners that will do presales and provide other marketing resources, and Acadia chartered to build and, if needed, operate Vblocks on premises for a customer. Who will you buy Vblocks from? Just about everyone except HP, IBM, and OracleSun.

Not to be outdone, the day following the VCE announcement, HP announced Converged Infrastructure (CI). CI is integrated server, networking, and storage resources too, but doesn't need a coalition. It takes most of what it needs from HP's own product lines. I say most because HP doesn't own a server virtualization platform--a key ingredient. Not to worry. Despite the fact that VMware is invested in Vblocks and Acadia, it likes CI too. And then of course there's the HyperV alternative out there...somewhere...

Its decidedly unclear at this point how successful these integrated compute stacks will be in an IT marketplace that's undergoing multiple transitions--from physical to virtual, from stove pipe to cloud, from decentralized to consolidated. The purveyors of integrated compute stacks are driven by a central belief: that your CIO wants to make the transition to virtual/cloud/consolidated simpler by wrapping up servers, switches, and storage arrays into one neat, pre-integrated package.

As storage pro, you may find that ironic. You may be old enough to remember when many of these same vendors were selling decentralized client server computing as the better, simpler way. Now the Three Musketeers--consolidation, centralization, and virtualization--are here to vanquish the complexity created by the-network-is-the-computer computing. Please shoot me if I even suggest that integrated compute stacks are the new mainframe.

So, as a storage professional slaving away within the bowels of corporate IT infrastructure, is the integrated compute stack about to change your life? I'd say yes if the powers that be like the Vblock concept. And if they like Vblocks then your choice of storage is EMC's. If CI wins, you get HP's flavor of the month array. Simple, right? And life could get simpler still when the IT operations group reorganizes around Vblocks. You may get to know the people in the server and network administration groups much better than you know them now as they start doing some of the things you do.

On the other hand, Vblocks and CIs may well be a tough sell in your organization. Cisco has yet to make its mark in the server world and buying a million-dollar anything from a coalition of vendors that want to run your critical applications on their collaborative platform is untried to say the least. But the biggest hurdle standing in the way of the integrated compute stack as it approaches your IT operations group may well be the following retort: "We just don't do things that way here."

For the last two decades, RAID (redundant array of inexpensive disks) controllers have ruled the storage world. RAID has been required for data protection in disk arrays. RAID schemes (RAID 0,1,6 10, etc.) reside on RAID controllers baked into disk arrays with many billions sold to date. But perhaps more important from the standpoint of making money, the RAID controller has also delivered differentiated value for storage vendors. Data copy and migration, snap shot, deduplication, and the list of controller-based functions goes on--all have been loaded on to the RAID controller.

It's becoming increasingly clear that the traditional RAID controller is coming to the end of its life cycle, at least within the enterprise data center. Types of applications now common to the Web 2.0 community are now populating the enterprise data center--applications that require scalability into the petabyte range. Traditional RAID controllers start to show their shortcomings at this scale level. Drive rebuild times elongate to the point where RAID data protection is no longer protection.

We can argue (and I have) over how much longer the RAID controller will survive. For sure, it's nowhere near dead and will continue on as the workhorse of the storage industry for some time. But its shortcomings are becoming increasingly obvious and are driving the creation of the next generation of storage devices. Indeed one of those devices is no "device" at all. Rather, it's software running on a collection of commodity servers and server-attached disk, both traditional and solid state disk. Think of this new "device" as software-defined storage where all of the functionality is defined and delivered in software. So as a user, when you buy a software-defined storage device, you're simply buying code. What you run it on is up to you.

MaxiScale is an interesting example of software-defined storage. MaxiScale's FLEX storage platform runs on standard servers with SATA disk, and uses standard Ethernet interconnections. It is implemented as clustered nodes--servers plus disk. I/O performance and capacity scales linearly as processing nodes and disk drives are added to the cluster.

So the storage value-delivery model is decidedly different here. You as the user buy software and essentially roll you own array. But what else is different here? First, while the RAID controller is gone, the absolute requirement to preserve data is not. Data protection is also implemented in software.

Second, the system assumes that individual nodes within the cluster will go off line or fail for one reason or another. That's OK. The FLEX storage cluster continues to function, perhaps at some degraded state for some period of time until the full cluster is restored. But the point is that once you power up the cluster, you can keep it running for years--decades if you want. Hardware is added and replaced without disruption. Software is upgraded without disruption. It's perpetual storage.

Third, FLEX is an expression of the state of the art in single or global namespace file system technology. It's this core technology that delivers the value-added storage services rather than the RAID controller.

MaxiScale is not alone in this emerging space. Other software-defined storage solutions include ParaScale's cloud storage software and Symantec's FileStore. Other traditional hardware and software players will follow with software-defined storage offerings in the coming months. Include database vendors in this space as well. Some will position their solutions as cloud storage, others as data protection and archival storage.

Will software defined storage replace traditional RAID storage? Not immediately. Not dramatically. But to me a new model is emerging. Scalability, hardware independence, and system longevity are the more compelling features when compared to traditional RAID-based storage arrays. But perhaps the most compelling feature will be an ability to buy big array performance and scalability at a fraction of the cost of big array RAID.

During the last two weeks we saw two acquisitions of relatively small purveyors of scalable file systems by big storage players. First, HP finally pulled in its partner IBRIX. Only days later, LSI made a surprise acquisition of ONStor. If both IBRIX and ONStor offer platforms upon which one can build scalable network attached storage (NAS), do these back-to-back deals indicate some sort of emerging trend? Yes and no. Yes it is in that, if you're a major NAS vendor and want to compete with NetApp who is readying GX8, scalability is now a must-have. But IBRIX extends capabilities HP already has whereas for LSI, ONStor represents their first ever venture into the NAS world.

Amazing is the amount of blogosphere and Twitter chatter that was generated by HP's announcement that it intended to acquire IBRIX for an undisclosed sum. No offense IBRIX people (all 53 of you), but you're not exactly a household name. It looks like HP is about to make you one however. And yes, you deserve all the attention you are getting, finally. You had a "next-gen" parallel file system before many knew they would even need one. You knew that Big Data users needed a file system that was system-agnostic and that would scale to the petabyte range. At the time however, they were in a niche-y place called high-performance computing (HPC). Now, Big Data users are cropping up everywhere. You count AOL, Caterpillar, Dreamworks, JP Morgan Chase, and Pixar among your 175 customers. Who knows where this cloud thing will take you.

Big is a relative term. In the storage world, what is big today will be table stakes tomorrow. The Petabyte-scale file system is becoming a must have for storage vendors. NetApp bought Spinnaker a while back. Sun developed ZFS. IBM has GPFS, and HP bought PolyServe last year two years ago but has chosen to position it in the Windows SQL Server space where it gets the most traction. IBRIX, with its many performance and data management capabilities, represents a much larger market opportunity to HP. And LSI has chosen to enter the NAS market as scalable from the get go.

IBRIX is headquartered in what was once a Honeywell Bull facility in Billerica, Mass. When they appeared in 2000 with a unique parallel file system called Fusion, the question was how to bring this to market? Who buys a parallel, scalable file system when file systems normally come bundled with or embedded in something else? IBRIX answered that question by forming remarketing relationships with big names: Dell, EMC, HP, and IBM who bundled/embedded IBRIX with their servers and storage. Dell and EMC packaged Fusion with PowerEdge servers and Clarrion storage, presenting the package to high-performance computing (HPC) customers. HP embedded Fusion in HP Blade and ProLiant server racks.

So what exactly does HP have planned for IBRIX? According to HP's Paul Perez, "HP will put the U in unified storage." OK, but that's a bit cryptic. Short term, HP will keep on keepin'-on with blade server/blade storage and scalable ProLiant/IBRIX NAS implementations. Longer term we may well see HP use IBRIX to approach cloud computing and archival storage opportunities.

Unified storage with a capitol "U" is a bit more of a challenge to understand. Typically the term has been applied to disk arrays that support fiber channel and Ethernet connectivity. HP likely means that kind of unification plus something more. IBRIX is typically used by its partners to create scale-out NAS subsystems using Fusion as the software engine that powers a NAS platform consisting of industry standard servers as the NAS front end, and SAN or direct-attached (DAS) RAID storage on the backend. As such, the combination presents scalable file storage to applications but uses block-based SAN or DAS storage. NAS is typically characterized as file storage, while SAN is block storage. It's a distinction that traditionally has had many application implications and ramifications. What HP's big U for Unified message may also be signaling is the introduction of a file/block converged storage product bundled with new hardware form factors sometime in the near future. For HP that likely means some combination of HP StorageWorks SANs, ProLiant rack-mount and blade servers, and ProCurve Ethernet switches powered by Fusion.

It's interesting that HP has chosen to announce a marriage now. After all, they've been dating for at least four years. But NetApp, after fussing like forever with the scalable file system it acquired from Spinnaker, is finally ready to go mainstream with it as ONTAP GX8. IBM is making more noise about GPFS. Then there's ZFS and its new owner--Oracle.

Which brings us to LSI and ONStor. LSI's Engenio Storage Group wasn't in NAS until now. It is in RAID arrays and storage virtualization. Now it's in scale-out NAS and NAS/SAN gateways too. Why? LSI/Engenio sells exclusively through original equipment manufacturers. IBM is a major reseller as is (was?) Sun. Dell is also in the mix. IBM's DS3000, 4000, and 5000 series arrays are all originally produced by LSI/Engenio.

But there is much repositioning going on among the big IT vendors these days. The future of Sun's hardware business is still a matter of debate in spite of Larry Ellison's assurance that Oracle will sell hardware too, and Dell is on record as in the hunt for companies worth buying. Is another storage acquisition possible for them? I think so. As a result, the number of large OEMs that LSI/Engenio can sell through is not growing and the future is unclear with regard to OEM sales of traditional RAID arrays via the big names in IT.

So NAS to LSI/Engenio represents new growth and possibly substantial growth if they can compete effectively. HP's acquisition of IBRIX potentially leaves something of a hole in the scale out NAS product lines of Dell, IBM, and EMC that LSI/Engenio. Dell and SGI may also need NAS/SAN gateways. IBM might like to have a second source for the NAS boxes they get from NetApp because ONStor is both scale out and scale up. And let's not count out Sun/Oracle either. Whereas IBRIX had established prominence in HPC computing, ONStor went after more mainstream applications and could be a better fit in that space for the big OEM partners.

So put LSI/Engenio on the list of buyers looking for things to buy as opposed to the other way around. They didn't just suddenly decide to write a $25 million check to the owners of ONStor after learning of HP/IBRIX. Their executives have assured me that they've been looking to add NAS to the portfolio for months. Market conditions and a desire to make a "we're here to stay" statement drove the timing of the ONStor deal.

Suddenly a somewhat dormant space in the storage world is erupting with activity. Why? The Big Data apps are here and they generate big system opportunities as well as big Unified storage opportunities.

IBM's storage group is now in the habit of making smorgasbord announcements. They'll take a look at their storage lineup--one that includes everything from SSD to tape, storage-related software and services--select the new stuff going on within each product development cycle they think is significant and therefore want to publicize, then bundle all of these separate announcements up in a wrapper ("Information Infrastructure" begets "Smart Planet")--and step up to the microphone.

And so it is with IBM's most recent storage table selection. They're now offering replication and deduplication for ProtecTIER, faster hardware and SSD support SAN Volume Controller, Thin Provisioning for DS8000 arrays, a new version of Tivoli Storage Manager, and numerous enhancements for XIV storage. Don't get me wrong. I'm not trying to belittle what they're doing. I would, however, like to observe that some of the more significant and interesting things can tend to get lost in the shuffle. XIV is a case in point.

XIV could well be a piece of computer history in the making because its guiding light, when at EMC, once took on and beat IBM at its own game. XIV was founded

in 2002 and emerged in 2005 with its first product called Nextra. It is an Israeli-based start-up and the place where Moshe Yanni landed after he left EMC. Yanni, known in the storage industry by just his first name ("mo-shay"), is the father of EMC's Symmetrix/DMX, the longest running disk array product family ever. He and his team created the MOSAIC 2000 storage architecture, which allowed EMC to update Symmetrix' disk, controller, and connectivity technologies more or less independently of one another. MOSAIC 2000 helped in a big way to establish the financial foundation that supported EMC's future expansions into content management, security, and virtualization.

As mentioned, XIV announced its first product, Nextra, four years ago--a next-generation disk array composed of clustered storage nodes. Shortly thereafter, XIV--and Moshe--were acquired by IBM. So here we have the father of Symmetrix, a product that allowed EMC to supplant IBM as the king of enterprise storage, now carrying the banner for the IBM storage team. Could EMC's former benefactor and acknowledged storage maven now become its biggest enterprise storage headache? Quite possibly.

By all accounts, Moshe doesn't kid around. Lore has it that he was once (and may still be) equipped with an Israeli fighter jet, and that after EMC bought the Clariion array along with the rest of Data General, which he saw as an internal competitor, he and his team built a small Symmetrix, stood it up outside his office door, and attached a sign to it that read "Clarrion killer." Lore also has it that his departure from EMC was literally an executive office glass-shattering event. In fact, there is a whole body of Moshe lore known mostly to storage industry cognoscenti--stories traded over beers. Who knows how much of it is fact? But one thing we all agree on: when the world of enterprise storage feels like a snake pit as it often does, you want Moshe on your side. Now he's on IBM's side.

IBM's acquisition of XIV raised more than a few industry eyebrows. A fiercely independent storage genius goes to work for the Big Blue marketing machine known as IBM? A clash of titans could be in the making. Well, so far so good. IBM dropped the Nextra label in favor of calling both the product and the company XIV, but has allowed XIV to field its own salesforce, as well as manage its own R&D budget and product development efforts. IBM is also promoting the XIV brand as an enterprise storage play, in spite of the fact that it also has its own internally-developed enterprise storage array line, the DS8000 series. IBM also allowed XIV to announce that it recently sold its 1,000th array, and that many of its new customers are former EMC Symmetrix/DMX customers.

Now that IBM has two enterprise disk arrays in the product portfolio, and two sales teams selling enterprise arrays to the same big systems customers, one could well wonder how IBM will differentiate going forward. Look to future announcements for clues. When the XIV acquisition was announced to storage analysts, IBM positioned XIV in "Web 2.0 storage"--that is, as something distinct from traditional data center storage where the DS8000 lived.

Well, ahem. Guys, nice try. We know a bit about Moshe and we don't think he's about to confine himself to a market subsegment. We now note that the most recent XIV announcement drops the Web 2.0 distinction and moves the DS8000 closer to the System z mainframe world--a place where XIV doesn't play because it doesn't support CKD disk formatting. It's not that XIV's engineers don't know how to do that. Moshe's Symmetrix started life as a mainframe-attached box. But there have to be some distinctions going forward. IBM mainframe customers get the DS8000 exclusively for now, but maybe not forever. And IBM is rumored to have one more DS8000 model to release later this year.

Watch future announcements for subtle shifts in messaging though as IBM will transition from DS8000 to XIV as its flagship enterprise storage array. The DS8000 is now called the "flagship mainframe array," while the XIV array has been promoted the "next generation storage" on IBM's most recent storage smorgasbord.

One more piece of Moshe lore--the name of the company XIV or the Roman numeral fourteen? What's the significance? Ask around and you may get conflicting answers. That's the nature of a legendary figure. The one I've settled on is this one: XIV stands for the fourteenth graduating class of Talpiot, an elite Israeli Defense Forces training program, of which Moshe and three other XIV executives were members.

Warning: This post is larded with acronyms. Sorry, it can't be helped.

Fusion-io recently announced SMLC flash memory. What is SMLC? To understand how Fusion-io coined this acronym (and it is all theirs), you need to first become familiar with two more--SLC for Single Level Cell and MLC for Multi Level Cell. Both refer to types of NAND flash memory, and both are hot items right now, but for different reasons. MLC NAND flash memory is relatively cheap, abundant, and commonly used in PCs, laptops, and mobile messaging devices. SLC is in comparison much more expensive, not as abundant, but is also hot because it is now being used to replace spinning disk in high-performance storage arrays by a growing list of vendors that now includes EMC, HP, and NetApp to name just three.

Fusion-io's SMLC is not actually a new addition to the flash memory family. (Fusion-io doesn't make flash.) SMLC is more like an attribute or feature, if you will, of a new solid-state disk controller that Fusion-io has introduced. What this new controller does is in essence give MLC flash the characteristics of SLC. Hence SMLC stands for Single-mode Multi-Level Cell.

Here's why that's significant. NAND flash has a shortcoming vs. disk as a storage medium in that you can't use it to store new bits of information over the same time period as disk. NAND flash begins to "wear out" faster than disk the more you write new data. This problem is more severe with MLC than SLC. At least one disk drive manufacturer says that this shortcoming plus some others makes NAND flash unsuitable for enterprise data center applications.

Undaunted, enterprise datacenter storage array vendors have gone ahead and implemented NAND flash-based SSDs and data center customers are buying them. These implementations use SLC because its longevity and reliability characteristics are better than those of MLC. However, as mentioned, SLC is more expensive, so the high-end array vendors are trading-off lower-priced SSD components for better performance, reliability, and longevity--a reasonable trade to make given the intended application. That makes SLC-based SSD less attractive. (Sorry, see death-by-acronym disclaimer above.)

Suppose however, that MLC could be given the more data center quality characteristics of SLC. The trade-off then goes away. That's the value proposition supporting Fusion-io's SMLC. The secret sauce here gets poured into Fusion-io's SSD controller. New controller-resident processes add the needed performance, reliability, and longevity to MLC, making it roughly equivalent to SLC in function, but at a far lower price point.

How significant could this introduction be? SSD for the corporate data center has always been hampered by price. In fact, even though data center SSD arrays are blazing fast, they have struggled to compete with disk arrays for decades. Even recently as customers are now looking seriously at the new flash-based SSD implementations from Intel, Texas Memory Systems, and others, price remains a stumbling block limiting SSD to address only the applications needing the best I/O performance. Lowering the price-point significantly, as Fusion-io does with this announcement, allows flash-based SSD to address a broader range of applications and replace disk in greater volume.

Beyond the threat this announcement poses to rotating disk, RAID (redundant array of inexpensive disks) as we know it is also threatened. Fusion-io's controllers can be clustered across physical boundaries to form server-internal storage networks. Data can be quickly mirrored across these boundaries yielding a solid-state storage subsystem with data protection that could compete on price with a RAID array.

To date, Fusion-io has OEM (original equipment manufacturer) relationships with Dell and HP. IBM has announced its Quicksilver project based on Fusion-io technology, but has not as yet delivered product bearing Quicksilver technology. SMLC will be available this quarter for Fusion-io's ioDrive and ioDrive Duo controllers. While these OEMs have yet to announce SMLC support, I expect the announcements to be forthcoming.

The Data Domain thrill ride ended this week when EMC upped its offer for DDUP to $33 per share (a whopping $2.1 billion dollars, or 111 times Data Domain's earnings), after which NetApp bailed out and went home.

Then came another surprise. Broadcom dropped its hostile takeover bid for Emulex--a prize it had been chasing for months. Hostile is almost an understatement for Emulex's reaction to Broadcom's entreaties. Harsh words and lawsuits were flying back and forth. But that one ended too when Emulex's board of directors unanimously rejected Broadcom's sweetened bid of $11 per share. Not even close to what we're worth, said Emulex. For Broadcom, this rejection was the final act.

Think we're done now for 2009? I think not. Stay tuned, because the storage industry has become a cauldron of acquisition activity. Yes NetApp lost its bid for Data Domain, but is likely not out of the hunt, if it wants to try again--for instance, other dedupe technology sources include ExaGrid, FalconStor, Quantum, and Sepaton. Nor can we say that Broadcom has suddenly lost interest in getting to the Fibre Channel over Ethernet "Go" square sooner rather than later just because they were stiff-armed by Emulex. There are other FCoE processing stacks out there to be had--via an acquisition of QLogic, for example. In other words, those were just warm-up events.

Joe Tucci, chairman, president, and CEO of EMC, has proven over the years to be a master of the technology acquisition, so whatever he's watching, I'm watching. More than a year ago I was at an EMC World event at which Tucci spent some quiet time with a gathering of industry analysts like me. He can be surprisingly candid in these sessions. During this one he disclosed that he was closely watching four technology areas he thought would produce significant heat within the next few years: virtualization, data deduplication, solid-state disk, and cloud computing. Now it looks like he was right on all four out of four.

The first one, virtualization, was no real surprise. He'd already acquired VMware, by far the brightest star in the virtualization firmament, and he bought it for what now appears to be a $635 million song. So EMC now owns the big server virtualization franchise, but what about storage virtualization? EMC really doesn't have that square covered yet, and rivals like HDS, IBM, and NetApp are pecking away at EMC's potential clout in this space. Will Tucci turn up the heat on internal development, or get out the checkbook? Hard to say, but FalconStor for example is potentially available; it should be remembered that FalconStor was once a storage virtualization play before it found a secure niche in data deduplication.

Speaking of which, the second area, data deduplication, was a bit of a surprise in that dedupe feels more like a feature than a product. Furthermore, EMC makes big money on storage, and deduplication makes big storage smaller. Why would EMC want to go there? No matter. EMC is about to pay billions of dollars for a company that sells disk-based data protection appliances competitively against EMC. Dedupe is the prize inside because, as a feature, data duplication can go just about anywhere--it's applicable to primary, backup, and archive storage--and it actually induces buyers to take on more disk and less tape. Clearly, Tucci wants EMC to own the dedupe opportunity (and DDUP for that matter) wherever it goes. Competitors will take note and respond. Given the current economy, storage efficiency is now the rage, and M&A activity around dedupe is far from over.

Solid-state disk was a big surprise at the time. SSD arrays once kicked around big data centers years ago, but never got a firm foothold. Moreover, in those bygone days of the early 1990s, EMC actually had an SSD array that it killed off when its Symmetrix disk array sales took off. Now solid state is back on EMC's radar screen because finally, and as a result of the ubiquity of NAND flash memory, it can be bought at a price point that compares favorably with rotating disk for applications that demand performance. Finally, SSD is hot. The companies potentially in play here are not the makers of NAND flash (like Intel and Samsung). The hot start-ups to watch are the developers of SSD controller technologies ( Fusion-io and WhipTail Technologies for example)--that is, the people who take the NAND flash and actually do something with it. The real work of managing NAND flash and presenting it as disk to an operating system gets done at the SSD controller level, and if Tucci is watching this space closely, he's not alone.

Seeing cloud storage on the list was once again a big surprise. That's another place the storage industry has once before gone and failed. We remember from the Web 1.0 era the meteoric rise and fall of many Internet-based storage service providers like Storage Networks and StorageWay. Even Enron had a horse in this race. But that, as it turns out, was cloud storage 1.0. Cloud storage 2.0 is big and growing, and again Tucci has proven himself to be on the mark. Indeed, EMC has already made an acquisition here--Mozy, once a cloud-based data backup service that now anchors EMC's Decho. EMC is also actively developing and marketing a cloud storage platform called Atmos, and has recently introduced Atmos onLine. Does EMC need more cloud storage? Only time will tell. And even if Tucci doesn't see a need, his competitors will. The potential cloud storage acquisition targets are many and varied, from cloud storage service plays like Nirvanix to cloud storage software plays like ParaScale.

With Broadcom's shadow away from its doorstep, Emulex now gets to prove that it really can spin FCoE chips into gold on its own. And while Data Domain didn't wind up with the suitor it wanted, Tucci is working hard at making DDUPers welcome in Hopkinton, Mass. Is that all for storage industry M&A activity this year? Not by a long shot.

Broadcom's persistent attempts to drag Emulex to the acquisition altar took another turn recently when Broadcom sweetened its offer from $9.25 per share to $11 per share of Emulex stock. Now would be a good time to stand back and do a quick reality check: Broadcom, tell us again why Emulex is now worth approximately $912 million--$148 million more than it was, say, a week ago.

In a May 5 press release announcing its first unsolicited offer, Broadcom's executives said, "Emulex has made it clear that it shares our view that the convergence of data and storage networking is the long-term future of enterprise networks. This combination would enable us to accelerate our efforts to bring this vision to our customers." While upping the ante on June 29, Broadcom amplified that statement. "Together, the talented employees of our two companies could accelerate the convergence of Ethernet and Fibre Channel. Broadcom's technology, scale, track record of execution, and highly successful history of acquisitions, along with Emulex's considerable strengths today would make a terrific combination for our combined employees and our customers."

Both statements point to the presumed convergence of data storage (Fibre Channel) and networking (Ethernet)--"presumed" because most large enterprise data centers have yet to go beyond the testing stage of this emerging technology. There is now a standard for this convergence, called Fibre Channel over Ethernet (FCoE), a protocol for attaching Fibre Channel-based storage devices to application host processors over a Gigabit Ethernet network. Emulex has been able to implement the FCoE protocol "stack" on a single chip and that's a major accomplishment worth serious money as Broadcom sees it. But at best, FCoE lives out on the bleeding edge of technology in the minds of data center administrators and other IT professionals.

At least one such person I follow, Scott Lowe, has raised questions that are likely reverberating within the enterprise IT community. In a blog item, he asks "Why deploy FCoE?" This is a serious question that he's not offering up as a lead-in to sell FCoE. And he may well be like many who, when first confronted with the idea of Fibre Channel-based data storage and networking convergence over Gigabit Ethernet, was persuaded that FCoE was the undeniable future. But now that some people within the IT community have had a chance to test-drive FCoE, doubts are emerging.

The simple fact is that no single information technology dominates the landscape for long. Mainframes had to make room for open systems and client/server. Linux ended the hegemony of Windows and the open systems primacy of Big Unix. And Ethernet will likely not be the only interconnect data center that administrators ever rely on. InfiniBand marches on like the Energizer Bunny, and Fibre Channel itself, even without the Ethernet assist, is still a long way from being at the end of its road map. We can't count iSCSI out yet either.

So, back to the Broadcom-woos-Emulex soap opera. Broadcom tells Emulex shareholders that they're crazy if they don't take the original deal. Emulex tells its shareholders not to listen to a company that used to be run by a man with a questionable past.

Unfazed, Broadcom ups the offer. Why? Because it wants that single-chip implementation of an FCoE network adapter. Emulex has it and is already claiming design wins with would-be OEM vendors. Broadcom wants into the converged data center network opportunity really badly, and it sees Emulex as the fastest way to get there, so much so that its executives are now willing to put up close to a billion dollars to get it rather than going through the cost and hassle of developing single-chip FCoE themselves. Up to now, the response from Emulex has been, "Go away, you wolf. We can get along by ourselves, thank you very much."

And that's where this standoff is now. As Emulex pauses to reflect on Broadcom's $11 per share offer, perhaps Broadcom might want to take a deep breath as well. While Emulex has other product opportunities working and has a 30-year history as a successful OEM vendor--all of which add value--Broadcom appears focused on Emulex's FCoE chip and the data center converged network thing. Data center fabric convergence a la FCoE is a major opportunity, no doubt, but pure Fibre Channel still has a considerable amount of gas left in its tank, iSCSI is getting a boost from VMware deployments, and we can't pretend not to hear the steadily advancing InfiniBand drum beat. The race to data center interconnect dominance has only just begun and may well wind up without a clear winner.

In the run-up to this most recent offer, Broadcom has steadily asserted that Emulex can't deliver value to shareholders beyond that which it already has. But Emulex has delivered product value for years within the context of other more stable storage networking technologies: Fibre Channel, iSCSI, and InfiniBand. FCoE is still an unknown quantity. Remember all the interoperability issues that plagued Fibre Channel years ago? Think FCoE is immune to them? Think again.

Yes, Cisco is out there hyping FCoE, and Cisco is a powerful force in the marketplace. But note that Cisco's director switch product recently lost significant market share to Brocade. Therefore, Broadcom can't depend on the power of Cisco to make the FCoE opportunity real. Data center storage administrators are a very conservative lot and have seen hype curves come and go. Reliability is king to them and FCoE has yet to prove itself on that score.

Broadcom is to be congratulated for upping their bid, not because Ethernet/Fibre Channel convergence is that big of an opportunity, but because the new offer is more reflective of the true value of Emulex as the sum of all of its parts--including the FCoE chip. The opportunity that Emulex represents remains a storage networking opportunity and not merely the dream of a converged data center fabric that may or may not be real.

Adaptec has announced the immediate availability of a Series 5Z Unified Serial (SATA/SAS) RAID controller family with "Zero Maintenance Cache Protection." It's designed to replace the current generation of RAID controllers that use lithium ion batteries to protect data in cache memory.

This announcement is significant for two reasons. First, its shows the expanding role of NAND flash. Recently, NAND flash has been associated with the second coming of solid state disk (SSD) in enterprise disk arrays. In spite of reliability issues (the more flash memory is written to, the less reliable it becomes over time), silicon vendors including Samsung and Intel now claim that flash can be made ready for prime-time data center applications by using a "wear leveling" process. Vendors like EMC and Sun have agreed by implementing flash as "Tier 0" storage within their arrays.

Adaptec is now using it for another application: NAND flash coupled with Supercapacitor technology to back up RAID controller cache in case of a system or power failure. RAID controllers use cache memory to accelerate I/O performance between the host system and disk. At any given moment during normal operations, there could be hundreds of write operations waiting to be transferred from cache to disk that. In the case of system or power failure, those could be lost. RAID controller vendors often deal with this exposure by including an optional battery power capability that keeps cache (i.e. RAM) powered-on for between 48 and 72 hours.

Here Adaptec does it a bit differently. It uses 4GB of NAND flash coupled with Supercaps to preserve I/O transactions stuck in cache. In case of a system or power failure, transactions in cache are destaged to NAND, to be "replayed" when normal functioning is returned. In this application, reduced reliability over time is not an issue since the NAND flash would be used in rare occasions, as a failure recovery mechanism.

The elimination of batteries to support cache backup is the second reason this announcement is significant. Engineers who design RAID controllers have kept up with the advance of battery technology over the years. Here, good batteries are those which are small and powerful. Consequently, many have preferred to use the lithium ion variety.

But there is some pain involved. Lithium ion batteries have to be "conditioned" before they are put into productive use. Their operating temperature has to be controlled. They also require periodic testing, as they have a limited life-span. Many fail completely after three years. These shortcomings require that users periodically test the batteries, a process that forces the removal of cache from the I/O path during testing. Finally, vendors who ship products with lithium ion batteries outside the US have to endure a border-crossing "red tape" process which they say can be a royal pain.

With delivery of the 5Z Series, Adaptec has replaced batteries altogether with Supercaps. Like batteries, Supercaps hold a charge over time but work from a different electrical principle. In the 5Z design they are used to power the data de-stage operation from cache to NAND flash storage. Unlike lithium ion batteries, they can be put into productive use immediately, don't need to be checked periodically, and won't get held up when crossing the border.

It's a breakthrough that other vendors of RAID controllers are also working toward. You can expect to see competitors release NAND/Supercap cache backup before the end of the year. For now, Adaptec stands alone. But it is likely that battery-backed cache is about to take a spot in the museum of computing history.

Much drama has ensued since NetApp announced the intended acquisition of Data Domain on May 20 for the whopping sum of $1.5 billion.

EMC countered with a $30-per-share offer valued at $1.8 billion. NetApp then raised its offer to $30 a share, valued at $1.9 billion. Data Domain essentially said, "Thank you, EMC, but we like the new NetApp offer more than yours." EMC then claimed that it had been unfairly shut out of the bidding process and appealed directly to Data Domain employees.

NetApp countered with a claim that EMC's potential acquisition of Data Domain would fail a federal regulatory review, a claim that EMC has rebutted as it considers shoveling more cash into the fire to make its proposal more attractive.

To its suitors, Data Domain is now reportedly worth $1.9 billion. To give you some perspective on that figure, Oracle recently agreed to acquire Sun Microsystems for $7.4 billion. A $1.9 billion acquisition would mean that Data Domain is now worth about 24 percent of that number, yet its 2008 revenues of $274 million are a tiny fraction of the $13 billion Sun took in sales revenue during 2008. Here's another relevant data point: EMC acquired VMware for a mere $635 million.

Deduplication is the storage world's new killer app. It's the great shrinking machine. Think of the old Steve Martin "let's get small" routine. It shrinks big data down to a small fraction of its original size--way more than is possible with the more common data compression routines. Why is that process now worth billions of dollars?

Most IT shops are moving away from using tape as their primary backup media in favor of disks. Deduping makes this migration economically viable by greatly reducing the backup data footprint on disk arrays by factor of 20 to 1, on average. You can't do that with tape. Nor can you get the input/output performance of disks from tape.

But that's not all that deduping does. It can be run against primary data storage streams to reduce the data footprint within expensive primary storage arrays. NetApp, among other vendors, supports this. Running it here may amount to the functional equivalent of buying another array, given the capacity that's saved as a result. When IT budgets are constrained, and storage is one of your top budget priorities, that's a big deal.

One can also dedupe archival storage, making the disk a repository for archival data that may need fast accessibility on a periodic basis--like when your corporate attorney needs to find exculpatory e-mails from three years ago and needs them yesterday.

So now everyone has to dedupe. Every major storage vendor, from EMC to Hewlett-Packard to IBM, now offers at least one dedupe option of the many that are now available, including the in-line and post-process variants. IBM, for example, offers four options.

In spite all its high-profile competition, Data Domain has been the acknowledged leader in integrating deduplication into the backup process. It offers disk-based deduplicated storage arrays for heterogeneous backup environments, and it leads all contenders in this space, in terms of market share, by a wide margin.

Does a leading position in a killer app justify a $1.9 billion valuation for a relatively unknown company mining a niche storage opportunity? Stay tuned. The executives at EMC and NetApp hate to lose, and EMC may yet win the heart of the fair maid named Data Domain.