In a climate where falling hardware margins have pushed many storage vendors into pursuing software-based technologies, Veritas Software Corp. is looking ahead to a world where Web services, grid computing, and intelligent storage architectures combine. Veritas CTO Paul Borrill outlined his vision for the future of enterprise storage.
Q: How do you position Veritas in the enterprise landscape?Borrill: One of the things that gives us a strong position in the storage software marketplace is we are heterogeneous across different systems backbones and across different hardware arrays. We provide the glue that connects all these systems together, which customers find is absolutely essential because the complexity of managing storage is just exponential. The pieces of glue that we provide are the high-availability solutions that provide you with the failover capabilities and the redundancy to keep systems running. A second one is this protection where you're really backing up data to prevent it from being lost in a disaster or accidentally deleted, and that's our family of backup products. Finally, I think the third leg on the stool in terms of the product family is our storage virtualization products. We find that the true value that Veritas presents is the ability to have one set of products that works across all the platforms. So I think Veritas is extraordinarily well positioned, and very often we find ourselves being the Switzerland in the industry. Companies come to us to broker agreements between each other [when] they can't talk to each other because they're competitors. It gives us a privileged position to [be a] facilitator [in] incubating new standards activities. For example, the new Blue Fin specification that came out, which has now gone to the SNIA [Storage Networking Industry Association], was initiated by Veritas. But we all had signed agreements to work with each other and not publicize it to any one company's particular advantage. This Blue Fin specification ... is now being heralded as the savior of SAN interoperability and the very thing that everybody's [been] waiting for a long time.
Q: How much influence do you have over heavyweights like EMC when it comes to setting standards agendas?Borrill: Hardware vendors are inevitably going to be pushing their own inimical interests over and above what the customers really need, which is true interoperability. And so the question I would ask about EMC (Corp.) would be, "Do you ever see that they can truly gain credibility in a heterogeneous environment, when it doesn't really help to sell their hardware?" Now some companies, IBM (Corp.) would be a good example, are able to overcome this immense internal cultural pressure to always prefer their own platform. And IBM Global Services, for example, have done a fabulous job of basically saying, "We have the computers, we have the software, we have the storage, but guess what? We don't have to sell all three."
Q: Standards are a requirement but they're also a result of built-in interoperability, and companies have their own agendas as to why there is a need for interoperability. Is it the cart before the horse a little bit here?Borrill: You could almost say that standards are an emergent behavior of the industry. It has to happen for the industry to fundamentally go to the next level of success.
Q: And we've seen that with things like iSCSI, for example.Borrill: It's difficult to say anything good or bad against iSCSI. I was one of the very first people to say that one of the things you'll see happening in the future is you'll see SCSI blocks going over IP, and I said that five years ago. For example, one of the questions I was asked [was], "If the blocks are going over IP, as opposed to going over fiber channel, what problem [are] they actually solving?" When you've got billions of elements to manage in the form of individual blocks, that's inevitably more complicated than managing fewer entities in the form of files. So we see that one of the ways of managing complexity is to go to the next, higher level of abstraction and files. You could argue that what users really want is files, they don't want blocks. It just happens to be that the only technologies we have available today to be able to help provide a single, unified view of storage is block virtualization. And that's been true for a very long time. And Veritas has been very successful at doing that.
Q: Companies like Microsoft (Corp.) are working to solve the issue of managing the complexity around structured and unstructured data. What's Veritas' approach?Borrill: One form of complexity is what I call the exponential form. Examples of that are interoperability issues. Interoperability requires an entire [infrastructure] chain to work. If you want to think about why this is an exponential, I'll give you an example. As the number of managed entities [grows], the number of files, the number of blocks grows. And what happens is you end up with this very poor form of complexity which current storage architectures have. It means that you need more administrators to make it work.
[As] the value of hardware is going down, software suddenly is becoming the darling of the industry and everyone else is now validating Veritas' software position. Everyone wants to get in the storage software business. And then what happens is you start to see there's a very insidious factor coming in; the systems and software are now just too complicated. Now we need a lot more people, administrators to manage what's going on. We saw this in the 1920s ... when someone said, "At the rate of new telephones being installed in the country today, everyone in the country is going to have to be a telephone operator within 10 years." That caused the invention of telephone switching. Well guess what, we need that kind of next-level technology invention so we don't need as many people to manage storage.
Q: Enterprises seem to be turning to hardware vendors for storage management software. They're avoiding mixed environments because of the complexity. What impact is this having on Veritas?Borrill: What's happening is CIOs are finding a tremendous pressure to reduce diversity in their environments. How do you do this? Well, you use the diversity in your system. You have one hardware vendor, you have one operating system, and so on.
Q: A single throat to choke?
Borrill: Exactly. But there is a counter pressure to that. The counter pressure is you can't get the best in value or the best in breed of products from any one vendor. What also happens is if you falter or you fall behind [as that one vendor], then what happens is you use your lock in with your customers to force your customer to continue buying what is now becoming an inferior product relative to the rest of the industry. And CIOs know that, too. EMC is desperately trying to reposition themselves as a software vendor. It's validating what we've been saying all along, that it's [all about building] open software across more heterogeneous platforms.
Q: We're coming back to the standards argument here again. What chance do we have of Veritas technology becoming a dominant standard?Borrill: We don't have the power as a single company to be able to set one standard. In fact, the interesting little secret inside Veritas is we need standards ourselves. How are we going to port our software and make all the interfaces work on all these different platforms? It's in our interest, actually, to try and encourage all of these companies to have a consistent interface, so that our job of filling in the white space in all this complexity ... between platforms becomes easier. So it turns out that it's easy for Veritas to adopt standards because it actually ... makes our life easier and we know that it helps customers. And so to us, it's a virtuous circle.
Q: As storage hardware becomes commoditized, what solution can you see for enterprises that are concerned about being overrun by a big services company?Borrill: We think there is an alternative to using more and more people. And that's where the industry needs to go. Veritas has an internal goal on our product development which is called zero administration. So in other words, the actual design goal for many of our products internally is to completely eliminate the need for administration internally. Now, are we going to get there overnight? Absolutely not. It's like a zero-defect quality program. If you don't aim for zero, you're not even going to get close. The only way that we're even going to get close is because we're going to make some quantum improvements in the way the architectures are put together.
Q: How far down the track of the zero admin goals have you gone?Borrill: I think over the next two to five years, you're going to see a complete solution. I haven't got any products I can announce to you today. All I can tell you is that we're thinking about this in a fundamental way. We're not thinking of this in terms of how we can incrementalize on top of our existing products. We're thinking about [what] would be the revolutionary thing that would truly create a solution in this area?
The CIO has a constant amount of dollars that are spent every year. What happens ... is all of a sudden [hardware spending] precipitously drops away. Then what happens is you find that software comes out as being the dominant [expense]. Then the administration costs suddenly dominate the whole picture. This can't go on; at some point in the future, the CIO is just going to hire more people and not buy any hardware or software. So you can see there's a wall here that you can't get past. So fundamentally we're going to have to invent switching, or the equivalent, to encapsulate and to conquer this complexity that users have.
Q: So how does this intersect with emerging autonomic or self-healing computing initiatives?Borrill: There are two answers to that. First of all, autonomic computing is absolutely trying to address the right problem. It's trying to address the issue of how can we make the system much more automatically adapt to changing environments. It could be adding more storage, it could be just reconfiguring what you have already because your business conditions have changed. The question is can you design it in such a way that creates an adaptive solution so that [fewer] human beings have to be involved whenever these responses are taken? And the answer is yes. What you have to do is take lots and lots of very simple processes, lots of very simple rules, and put them in individual systems, treat them as a peer-to-peer system, and then create a solution which has an emergent behavior for what you're actually looking for. When it comes to storage, you're looking for resilience, you're looking for low latency and performance. These are the kind of emergent behaviors that you'd like to come out of a large and growing and scalable storage system. Now you've got an architecture which can fundamentally adapt without human beings having to be involved.
Q: How will Web services affect your goals and strategy?Borrill: I don't think it will affect our goals and strategy at all. Web services were taken into account as we were developing our strategy over the last two years. We've seen Web services coming for a while. To us, Web services are the ultimate distributed computing platform. It creates an emergent behavior where you've got things that can now talk to each other that never existed before. And Veritas is able to support that because all the systems need storage. And storage needs to move. So you're going to end up seeing more and more of a convergence between storage and networking, which becomes a foundation for Web services in the future.
Q: There are plenty of growing pains around Web services standards right now. Are we going to see Veritas become involved in that debate?Borrill: I think you're going to see us become more and more involved with partnerships with other companies in supporting Web services. We will retain our Switzerland-like position, but if you think about this, solving a distributed computing problem is an incredibly difficult problem. Web services is the first realization that what you needed to do is create an invulnerable structure.
Q: What do you see as the role of the CTO in terms of evangelizing your interoperability goals?Borrill: [A significant] part of the CTO's job is looking at those multiple facets of how the technology and the industry might evolve and saying, "I am willing to place a bet on those three or those five different options." I know that only one or two of them are going to succeed, but I want to make sure ... that I have all those options lined up so that when the time comes, I can say, "Okay, grab that technology, put it into the product group, get it out on the market." Now whether that's a Web services technology or whether it's a new storage technology or whether it's a clustering technology, those are the options we can use to create the opportunity for executives to make decisions in the future. So this is why it's very hard to say what we're coming out with next, because we may have some technologies on the way, [or] we may decide to shelve it, we may decide to cancel it, we may decide to continue it as an advanced development, or we may decide to say, "Take that, productize it now."
Q: What is your perspective on the emergence of grid computing?Borrill: Grid computing is something that's coming up fast on everyone's radar screen. We've been looking at it now for about three years. Veritas is a member of the Global Grid Forum, and we're looking at that as a potential standards activity that we participate in. I think it might turn into real commercial potential in a few years. We have people inside the company who are focused on that. Now, let me give you an example of where grid computing is creating congruence with what's happening in another part of the market. Grid computing came out of the idea that you had all these big machines lying around that had tons of [under-utilized] resources. Then there is [the emergence of] server blades. All of a sudden you've got hundreds or thousands of very small, very powerful machines. So what's common about these two [technologies]? First of all, you want to be able to bootstrap these things. You want to be able to take an image and you want to be able to clone that across 1,000 blades at once, instead of having to go to each individual blade and do an install. So what happens is you need to be able to make sure that the images you're loading are adaptive in the sense that they can handle [hardware errors]. Making an application work on the next blade with this much glue is really where hard problems are. Exactly the same problem that's mirrored in grid computing is seen in how you provision blades. So I'm seeing a collision course between two disparate parts of the industry headed for the same inevitable point.
Q: A blade can be seen as a client as well in a peer-to-peer model.Borrill: Absolutely. Or it can be seen in the peer. One of the things that I'd like to leave you with on this idea of grid computing is that Veritas is only solving part of the problem. Grid computing is a very broad, very big almost grand challenge kind of problem. And we're going to let the big boys fight over that one. What we're going to do is we're going to focus on just storage. We're going to have the best possible foundation for storage that can be used across blades, across the grid, and we're going to focus on that narrow area where we have absolute industry superiority in our core competence. And that's how Veritas is going to provide support for the grid.