Mike Wall, GM of Intel's Storage Components Division

Intel Corp.'s (Nasdaq: INTC) campaign to "PC-ify" the storage industry is already well underway, and one of the technologies it sees at the flash point of the transition to ubiquitous storage networking is iSCSI.

But the iSCSI revolution has stalled. Whereas the block-level IP storage networking protocol was once supposed to easily depose Fibre Channel SANs, it's fallen short of initial expectations, as even its most ardent supporters acknowledge (see IP SANs: Coming of Age).

That said, Intel is still a believer in iSCSI's inevitable ascendancy. Mike Wall, general manager of Intel's Storage Components Division (SCD), says it's just a question of when, not if: "Ninety percent of the time, if not more, it always takes a year or two longer for these new technologies to be adopted by the marketplace. But this is one that I'm sure will happen."

In fact, Intel has established a close partnership with Emulex Corp. (NYSE: ELX), one of the dominant suppliers of Fibre Channel host bus adapters a pairing that promises to provide the foundation for Emulex's transition to iSCSI (see Intel & Emulex: iSCSI Tag Team?).

Intel's SCD delivers storage system nuts and bolts, including I/O processors, PCI bridges, and RAID products for server and storage systems. The division's primary competitor is Broadcom Corp.'s (Nasdaq: BRCM) ServerWorks subsidiary.The idea, says Wall, is to apply the company's model of high-volume silicon production to storage. Storage platform vendors historically have developed their own proprietary components from scratch. "That's an extremely expensive way of doing business," he says. Instead, Intel proposes that storage companies use its standard, commodity building blocks, thereby lowering the overall cost of systems and allowing them to focus on high-level features.

Since joining Intel in 1992, Wall has held several positions, including senior director for SCD's Server/Storage Operation, general manager of its supercomputer systems business unit, and marketing director for Intel's Enterprise Server Group. Prior to joining Intel, Wall was a general manager at supercomputing company Cray Inc. (Nasdaq: CRAY), and also held various marketing management positions at IBM Corp. (NYSE: IBM).

Wall recently spoke with Byte and Switch US Editor Todd Spangler about the promise of iSCSI, the challenges of bringing storage networking to the masses, and the continued growth of Fibre Channel. Click on the links below to read the full interview:

Byte and Switch: Can you give us an overview of Intel's storage components business?

Wall: For the past two years, we at Intel have taken the storage industry a lot more seriously than we have in the past. Computing is a three-legged stool of computing, communications, and storage. So storage provides a great opportunity for Intel. We believe that this opportunity is really in its infancy in terms of the volume-shipped perspective... And we want to participate in that by providing silicon building blocks so [storage systems] can get to market more quickly and less expensively. That's an area that ties very well into Intel's core competencies of silicon development and manufacturing.Byte and Switch: How big is your group? Can you quantify it in terms of people or investment?

Wall: I can't talk financials. The storage components division has been around for about six years now. We have on the order of 400 people spread across multiple locations. We do a family of I/O processors that were formerly based on the 960 technology; all the new ones are based on the XScale microarchitecture. We also have a family of PCI bridges and RAID products. That's in our division. But within Intel there is storage technology being developed across five or six divisions. A lot of the business and technology discussions inside of Intel we look at that in a matrix perspective. So in the LAN Access Division, they're doing iSCSI, silicon iSCSI, TCP/IP offload engines. We're doing optical devices. We have the Embedded IA division focused on supplying Pentium processing to external storage array providers. We have the Intel Architecture Labs driving new initiatives and looking at new ways to drive storage applications, whether it's object-based storage or whatever in the future.

Byte and Switch: What technologies specifically are you focusing on now? I know iSCSI has seen its share of hype and heartbreak.

Wall: There's no mystery that iSCSI is a great opportunity. But I wouldn't be the first one to tell you that we're not as far along today as we hoped we would be when we looked at this several years ago. That said, we are still extremely bullish on iSCSI, and we think that iSCSI – along with a technology like Serial ATA – are providing certain transitions to the marketplace to facilitate that growth of storage platforms in the marketplace. [Ed. note: At least he didn't say "paradigm shift."]

A few developments lately have laid the foundation for the adoption of iSCSI. We finalized the iSCSI standard, thereby eliminating any doubts about the stability of the standard, and second, Microsoft Corp. [Nasdaq: MSFT] announced support for iSCSI. Those two things were clouds that hung over the iSCSI initiative and now it makes it more possible to develop iSCSI devices with a greater deal of confidence. [See iSCSI Gets Go-Ahead and Microsoft to Unleash iSCSI.]Byte and Switch: What needs to happen for iSCSI to really grow?

Wall: A few things are actually falling into place. But there are some areas that need to be developed with respect to iSCSI targets. You have a lot of things going on on the host side, but on the target side we need to get those devices developed sooner. So SCD, our division, is providing the silicon today to companies like LSI Logic Corp. [NYSE: LSI] and Adaptec Inc. [Nasdaq: ADPT] so that they can get target-side devices to the market more quickly, and we're working closely with them. Intel is working closely with Network Appliance Inc. [Nasdaq: NTAP] on getting some advanced iSCSI target technology into the marketplace. We expect to see more of these products over the next 12 months. [See NetApp's IP SAN Wins a Fan, Adaptec Details IP SAN Betas, LSI Lines Up iSCSI Mates, and NetApp Blitzes on iSCSI.] We also see several vendors developing RAID arrays using Serial ATA with iSCSI. We think that's an inexpensive approach to external storage. And in some cases that will be deemed "good enough," even though it may not be as feature-rich as a Fibre Channel environment.

Byte and Switch: What do you mean by "good enough"?

Wall: "Good enough" is an expression we've been using here more recently. A small office, home office, or departmental system may not need a predominance of high-end features and high availability that someone in a Fortune 1000 enterprise would require for a storage subsystem. And so, if you can just get a bunch of inexpensive drives and perhaps lower-bandwidth networking environments, but it's good enough for your little company of 20 people, then we think systems like that will sell in volume. ... So we think inexpensive systems based on Intel technology will be good enough for now and will get more sophisticated over time.

Byte and Switch: What are some of your biggest challenges?Wall: As the industry goes through these transitions from ATA to Serial ATA, from perhaps Fibre Channel to iSCSI, from PCI to PCI Express, and SCSI to Serial Attached SCSI – I just rattled off four from the top of my head – we plan on being at the leading edge of a lot of these technology transitions. So our biggest challenge is getting the end users to rapidly adopt these technologies. It's getting the ecosystem and the actual end customers of these products to embrace and adopt these technologies sooner rather than later.

NEXT: Volume, Volume, Volume

Byte and Switch: How much cost do technologies like iSCSI or Serial ATA remove from the equation?

Wall: It's hard to answer that. But the history in the storage industry is similar to the development of mainframes and servers. Whether I'm providing a low-end box or a high-end storage box, I build all my own ASICs [application-specific integrated circuits] from scratch. That's very expensive. There are teams of hardware engineers; you have to fab your own stuff; and you're not getting the volume out of it. That's an extremely expensive way of doing business. That's the way the mainframe business used to work.

Now, Intel in servers – and what I'm telling you is, Intel in storage – is providing I/O processing and Intel Architecture processing to alleviate the need for storage platform providers to have to do their own ASIC development, so they can focus more on the software and firmware features, they can use the building blocks we provide, and therefore they can still maintain a very, very good profit model but they're getting a lot of their basic core hardware technology at less expense – which will be good for the end user.Byte and Switch: Why is Intel investing in a company like RASilient System? [See Rasilient Rustles Up Cash.]

Wall: That ties into what we're talking about. We're looking to plant seed funding into the market to develop these low-cost platforms using these new technology. You take RASilient: They're looking to deliver a low-cost storage platform based on iSCSI and Serial ATA, and so we're providing them with seed funding to help them accomplish that. We also made an investment in Wasabi Systems Inc. They have a real-time operating system based on NetBSD and are looking at some iSCSI target capabilities. We've also invested in DataCore Software Corp.; they're doing storage virtualization. [See Intel Takes Stake in Wasabi.]

Byte and Switch: So you're looking at these technologies – initially anyway – as targeting a low-end market because you think that's where the traction is?

Wall: It's volume. It's been Intel's operating model for a long time. The volume economics kicks in, and we can afford to continue to churn out more silicon, better, cheaper, and faster, and then that volume spawns innovation from companies that perhaps could not even think about it in the past, and I just mentioned three small companies that are trying to innovate and get some traction. Then it just keeps feeding itself.

NEXT: Fibre PowerByte and Switch: What about applying the Intel model to Fibre Channel? In other words, making it a "commodity" technology... although I know Intel doesn't have a heritage in Fibre Channel.

Wall: Well, we are constantly looking at the different types of storage networking technologies that are out there. Obviously Fibre Channel is a very, very well-established technology. We do not expect it to go away anytime soon. In fact, we see Fibre Channel growing significantly throughout the remainder of this decade. We believe that it will coexist with iSCSI. We expect iSCSI to grow. So what we have to do is support both.

We actually have Fibre Channel customers today. We provide our I/O processors to some Fibre Channel companies. We're working with those companies to deliver better, higher-performance products at a better price point. And so at the same time, we're driving as fast as we can on iSCSI. And the smart Fibre Channel customers are working on iSCSI implementations so they can be part of the transition in the marketplace as it happens, and they can manage that transition without losing business. A lot of their management software stacks – software, firmware, APIs – run on our I/O silicon.

Byte and Switch: Who are those Fibre Channel customers?

Wall: Probably the most visible one is Emulex Corp. [NYSE: ELX]. There are constantly developments happening that will help bring more value to end customers. And so some of these things are crystallizing right now.Byte and Switch: But you develop technologies that potentially compete with some of your OEMs. For example, Emulex has had its own iSCSI initiative in the works for some time [via its acquisition of Giganet]. Isn't that competitive with what you're offering?

Wall: In this environment, there's "coopetition" all over the place. Yeah, Emulex has their own iSCSI agenda, and quite frankly that could be perceived to be competitive with LAD [Intel's LAN Access Division], although they're actually using our XScale technology to implement that moving forward. Their technology has to enable their current users of their Fibre Channel technology to make the transition seamlessly, so their approach to how they tackle the marketplace is a little bit different from LAD, which is trying to embed iSCSI technology at the NIC [network interface card] level. So I won't argue with the point that there is at times competition between us and our customers, but we can give you probably hundreds of examples of that.

Byte and Switch: What's going on with Intel's TCP offload engine ASIC? [See Intel Wiggles Its TOE.]

Wall: Like anyone else in this industry, we're going to have 1-Gig and 10-Gig offerings. Within Intel, there are products being developed to accelerate the offload. I can't really talk about specific unannounced products, but rest assured we will have these technologies and we will have the building blocks necessary to maintain a leadership position. There may be different ways of tackling these opportunities based on a different set of product requirements, so we work closely with LAD, and we also work closely with companies like Emulex, LSI, and Adaptec so they can provide leading-edge offload technology and get their customers where they need to get their customers.

A lot of it depends on what type of transition their customers need to make. Some of these customers are more at the high end, and they're used to a Fibre Channel environment, and a very feature-rich environment, then they need to bring all the robustness and richness of the Fibre Channel feature set into iSCSI. So there's a certain way of designing and developing those products to enable that to happen. On the other end of the spectrum, there are those that have never done any storage networking – iSCSI's a new thing. They don't have any legacy to worry about. So you can tackle that in a different way. There are different alternatives that are being implemented and supported by Intel.NEXT: Is iSCSI Disruptive?

Byte and Switch: You said you see Fibre Channel growing through the end of the decade. Do you even look at iSCSI as being a "disruptive" technology, then, or is that not how you see it now?

Wall: Well, you know, these are great questions. And there are times when a bunch of us sit around the table and you wish you had a crystal ball. But we don't. I think there were some unrealistic expectations set several years ago, where everyone thought iSCSI would be disruptive and unseat Fibre Channel. And that's just not happening. I think the more realistic way of looking at this today is that they're complementary to each other. One of the proof points in that is what Cisco Systems Inc. [Nasdaq: CSCO] has done. At one point, when they first announced [a storage networking strategy] it was exclusively iSCSI; now it's more of a hybrid with Fibre Channel and iSCSI. I think over time things may evolve to where iSCSI will continue to grow perhaps at the expense of Fibre Channel, but this, I can assure you, is not something that's going to happen overnight. Until we get to 10-Gig. You know, 10-Gig is probably a technology point where some of that acceleration of the transition could be enabled.

Byte and Switch: But until then, you think Fibre Channel is very well entrenched.

Wall: In the enterprise, if you're in a Fortune 1000 environment you're surely not going to walk away from your Fibre Channel infrastructure. What you'll do is start to dabble in iSCSI, you might put in a switch that enables Fibre Channel and iSCSI. But in an enterprise, it's going to be a lot slower. Whereas we believe at Intel that actually a new opportunity in the small office/home office environment that really doesn't have a Fibre Channel infrastructure today, and they may just start with a low-cost storage platform based on iSCSI.Byte and Switch: So basically it's a brand-new market.

Wall: Yes.

Byte and Switch: And so you have all the challenges associated with creating a new market – educational issues and developing a channel, right?

Wall: Absolutely. What I said earlier was that iSCSI didn't achieve the expectations any of us thought a few years ago. But can you name a new technology or technology transition that did meet its timing expectations? I can't think of one. Ninety percent of the time, if not more, it always takes a year or two longer for these new technologies to be adopted by the marketplace. But this is one that I'm sure will happen.

Byte and Switch: Why are you so sure this is going to happen?Wall: Because I haven't been anywhere on this planet where people haven't been interested in getting behind iSCSI, and are working on products. There really aren't a lot of alternatives out there. And there are real products available now. So it's not like a dream or a concept. There seems to be universal support for this, it's related to technology that exists today that's pervasive across the planet – in Ethernet – so it's just a matter of time... But today, iSCSI does not have the capability of Fibre Channel and it will not for the foreseeable future.

Byte and Switch: In terms of what?

Wall: As far as bandwidth today, and even when it achieves the bandwidth perhaps in the 10-Gig timeframe, you have to look at robustness of the environment from a software feature set – firmware, software, the application environment. These are things that have been in development for a decade now with Fibre Channel and are well integrated into the fabric – no pun intended [ed. note: none taken] – of all large enterprise customers. And to displace that is a nontrivial undertaking.

— Todd Spangler, US Editor, Byte and Switch