The looming SAN storm in the SMB market: Part 3; Continuation of the roundtable discussion on the emerging SAN market for small to medium-sized businesses sponsored by Computer Associates, Dell, Emulex, Intel and Microsoft.

Marrone-Hurley: At this time, we'd like to open up the discussion and hear some questions from the audience.

Questioner 1: I'm from Gartner. I like this idea of ease of use. I'm still a little puzzled as to where is the central coordination of this effort to make these SANs easier to use? I hear from Dell, I hear from CA, I hear from Microsoft. Where's the nerve center of that activity? How is this coordinated?

Lofgren: Do you mean organizationally? Do you mean between the vendors?

Q1: Well, there's something that needs to be accomplished. And, by the way, I like my external USB drive ...

[laughter]

I think we know that one of the things that needs to be done is make SANs easier to install, and easier to use. And everybody's saying Dell's doing something, CA's doing something. We know Microsoft's doing something. But is there any coordinated center where all these people are working together effectively to get that done, if everybody has to contribute?

Padovani: At the end of the day, providers like Dell and other solution providers, our competitors out in the market, we have to coordinate and present that solution to the customer. We do a lot of testing with customers to see what solutions work. All of our suppliers also have done a lot of ease of use testing, and we work together to say what have we found--

Q1: So you're the center of this?

[laughter]

Padovani: There are a lot of centers. Claude said he uses the channel. And Mike has said that Dell is a channel. We own the customer and drive the solution to the customer. So ultimately we're liable for how that solution's presented to the customer. As far as ease of use, if we want to improve the experience of the customer, to lower the support cost, we have to make that ease of use and drive that ease of use across our partners. So it's Dell and it's all of our competition out there that are trying to use that as a means of differentiation in the marketplace. We want the best experience for our customers compared to our competition.

Q1: I understand that. But there's nothing like a solutions forum, or something that brings together people, who would otherwise compete, to work together.

Padovani: That would be SNIA, SMI-S, the whole platform interoperability effort.

Q1: But does SNIA have something directed towards this activity? I'm talking about the improved installation and operation feature.

Marrone-Hurley: Do we think that SNIA actually should have a SMB focus group, like a design guide, that determines what feature set they need. Do you do it in a coordinated fashion?

Smith: I don't think we would defer to a standards organization to define the market needs and specific features. Standards are needed--they're the frameworks that we need to communicate--and SMI-S is a great example. It's the first example we have of providing a management framework where host bus adapters, switches and arrays can all talk collectively through the same language, if you will. But defining specific features is another area all together.

Lorenson: That question is a very good question. The reason it may be hard to answer is because there are many pieces to the puzzle. From our point of view, that means we have to create partnerships with companies like Emulex, like Brocade, Cisco--that, traditionally, we did not really partner with. We have to open our roadmap with these people; we have to share what we're doing in terms of strategy. We have a partner group at Microsoft, specifically dedicated to storage, that didn't exist two years ago. Now, at least we can share where we are going and what our philosophies are to make these products easier to use. That's on the Fibre Channel side. On the iSCSI side, we took a bit more of a leadership position, and we are actually working with our partners on deployment and best practices for how iSCSI should be deployed. So it becomes a much more white-paper canvas to deploy iSCSI SANs because we worked from the get-go with all these iSCSI partners. So when you talk about the "central room" or "operation room" where it is being done, I want to say that we are trying very hard to have all these storage partners in line--as far as the Windows Server platform is concerned. It's a hard question to answer because there are many pieces of the puzzles, and some of us--Dell and Microsoft--often have to assemble more of the pieces.

Padovani: I think it's a competitive advantage. That's why we may not get to some type of central collaborative area. But I don't think it's necessarily important, because as long as you get the underlying technologies interoperating together, then somebody's always going to improve the GUI, somebody's always going to improve the look and the deployment characteristics of it. That, just like cost, is a competitive advantage.

Performance is a competitive advantage. Ease of use for the small business. Those three are all selection criteria for small business. So I almost don't think you want to get a group of companies together and say, let's come up with a standard.

Wall: Historically, setting the standards enabled this cycle of innovation. Standards are the key, the trigger point, to allow a lot of companies to play in an area where maybe only a few did before. So it's key to the industry organizations to get the standards set in the marketplace. Once that happens, and that is happening right now--which is why we're talking about this small and medium business opportunity--then that cycle of innovation happens. Maybe this is a simplistic view, but then capitalism and competition is going to be the fastest path to getting these end users the simple, easy to use things that we're talking about today. Not bodies of people that are going to play together and agree to things. It's going to be Microsoft wanting to win in the marketplace and therefore knowing what they have to do to satisfy and make their product more competitive than others. But standards are the key to enabling that competition to take place.

Questioner 2: I'm from IDC. Many of you have said that part of the secret of getting SANs into the SMB area is simplifying. Essentially, what you've all said is they only need a percent of the function that you're currently supplying into the enterprise. Can you give us some idea of the things you're currently supplying into the enterprise that you don't think SMBs needs?

Lofgren: One quick example is being able to take a look at a lot of performance metrics, such as real detailed information on I/O performance. I would doubt if a large percentage of the SMB market users need to know all this nitty-gritty, detailed, granular information. Now, they do need to know some things. Maybe they need to figure out that there's something there that's a bottleneck for them, but I don't believe that they're going to need those "137 different things." Those things just get in the way for them. Now again, they need to know if there's a problem. But looking at what the cache hit ratio is of this or that, for the most part, they're not going to be interested in. They're never going to look at it. They don't have the time because they're not just looking at storage: storage is kind of a part-time job for them. And all they really want to know is if there is a problem. In many cases, what they tend to do is solve problems through hardware (which may not necessarily be the best way of doing things). But given the time that they have, that they can dedicate towards managing their storage environment--that kind of detailed information is just not as valuable as it is to somebody in the data center who's managing several hundred terabytes where it really makes a big difference in their transaction environments.

Marrone-Hurley: Mike, from the Emulex perspective?

Smith: Well, that question affects us big time. For example, the Fibre Channel standard says things like you can support up to.... I don't know how many million nodes, but it's multiple millions of nodes. I would imagine most SMB environments don't need to support a couple of million nodes, right? And supporting just that capability requires us to store that data somewhere and there's memory implications. The Fibre Channel standard says you shall transfer data up to 10 kilometers. Well, I'm not sure many SMB environments need to do that either. There are implications to feature sets that drive data buffering that add memory costs, for example. We have environments at very large-scale computer centers that you would consider mission critical--business critical--where transactions per second/per dollar is the metric, where performance is really, really important. And in smaller environments, really banging out hundreds of thousands of transactions per second may not be a life and death kind of situation. So you can scale that back, for example, and reduce the cost and complexity of your products. The notion of how many concurrent transactions ... If, for example, you have an airline reservation system, you may want to support thousands of phone calls coming in. If you're a smaller business you may not need that capability.

So there are lots of areas where you can look at this and say "you know, if I was going to build a ground-up product for the SMB, I can draw a box around a portion of these and do a great job for that SMB--and they won't miss a thing." And that's not even talking about the notion of tuning the performance and tuning the environment, which adds the type of complexity that a data center manager just loves. "Wow, I can tune this driver to optimize the performance of an Oracle database working with this particular storage spec." That's a great feature for an enterprise. It's a horrible feature for an SMB. It asks questions that they don't even understand, let alone need to answer.

So there's a software aspect to it, there's a hardware aspect to it. And through that pipeline there's a support aspect to it. But the more you can hide and eliminate the "don't-cares," the more you can simplify your support. That's a big piece of this problem. It's the whole chain of costs, not just the products costs you have to consider.

Questioner 3: I'm a freelance writer doing a piece on this for Small Business Computing. I've got two questions. What is the smallest company anybody knows of that actually has a SAN? And what are the minimum technology infrastructure and storage needs that they realistically need to have in order to even consider a SAN, no matter how simple it is?

Lorenson: I have a customer in mind: Two servers, direct attached storage, moved to a new infrastructure. He went from a non-Windows operating system to Windows. Changed his workload to six servers, 120 users, and he used iSCSI for his SAN. So there's a very specific example of a small company that went from direct attached to a SAN. In this case, he used iSCSI because he had some Cisco switch laying around in his company that he could use. He didn't need to buy any cards for his Windows Server. So in this case--and I'm very familiar with the customer--they went to iSCSI.

Q3: And the second part, the minimum technology infrastructure you'd need to have in place, and how much storage--realistically, the minimum--before you'd even consider a SAN?

Lofgren: There's a break-even point where, from a cost perspective, you're actually seeing the benefits: but it's a moving target because you've got the hardware costs always coming down. As we get better with management, it moves. I know where it used to be a year or so ago. I know the number I used to use personally. But now I'd hesitate to even say it. It also depends on the user. An industry average can be a little dangerous sometimes because it's not representative, especially in a marketplace such as this where it's a question of what are you're really trying to do. So I think it's difficult to say.

Padovani: One of the things to consider is "What are the availability requirements for the customer?" Because I may only have 10 gigabytes: but if I have 10 gigabytes on a single hard drive and that's inside a server, and the server goes down, then I've lost the data. If you stick it on a SAN, even 10 gigabytes is a good enough because you can put that on a RAID set, you can be protecting the data and you can have it be highly available, so if that server dies, you can add another server and map it to that data. So I don't think there's necessarily a tasking threshold. I saw in one magazine, a customer consolidated 50 gigabytes. So it really doesn't have to be a size threshold. There's performance, availability, those are part of this market.

Lofgren: Raw capacity for one thing; but how quickly they grow their storage and how much of a pain it is to add drives and take the environment offline to do that is a very, very important aspect. Even if they start at 50 Gig, but they need to add 25 Gig every month, they're going to be disturbed a lot in their environment.

Marrone-Hurley: It can't be a capacity question. I'm aware of a customer that has 100 terabytes on a SAN, but then another half a petabyte that's still DAS. So it's more about the issues. Look at what the SAN brings to the table. It lowers your total cost of ownership. It enables you to share data. It enables you to more effectively manage your environment. So you have to take those types of things into consideration to say, what are the cost savings and what are the administrative savings and everything else that someone's going to realize by going to a networked environment versus being in a DAS environment?

Questioner 4: Since all of you brought up the issue of cost, and implied that Fibre Channel has had price points that are horrible to SMBs, let's talk about that. Emulex has a $500 HBA. Hewlett Packard sells Brocade switches for $5000, $625 a port. The people you're talking about, SMBs, they're used to buying snap servers for $995, NICs for under $100. Is that really affordable? Is Fibre Channel at an affordable point right now to SMBs?

Smith: I've talked to people in that situation. If you're in an environment that's got 15 servers and your IT guy starts backing up at ten o'clock at night when the last guy goes home, and he's not done when the first guy comes in the next morning, you've got a problem. If you've got applications running on a server that you can't fit any more disks into and you've got another server right next to it that's got five spare drives, you've got a problem. That's what's driving these guys to SANs. And for them, the math works perfectly. "Help me out of this problem. I need to centralize this and solve this specific business problem." It's not going to matter to the guy who has one server and can use USB. That goes back to the "shades of gray" comment I made earlier.

Padovani: The $995 snap server has two ATA disks in it, or less. There's not much data availability to it. To that point, it's what level of protection does a customer want for the data. And there are some customers that are going to realize the value that they need for disaster recovery and for the availability of the data.

Q4: So what you really have to be able to do is, when you're selling these people on it being cost justified, put a dollar benefit on the backup solution itself.

Padovani: Two things. The cost will continue to drive down. And I think we've all said that as standards are adopted, we'll continue to see costs being driven down. We'll see lower cost solutions coming on the market continuously. That's always happened. And then there's also the communication. Some small businesses see the benefit already and don't have to be educated on it. Some customers have a problem and will need someone who can communicate the value of one solution versus another.

Questioner 5: Earlier on, everybody was saying "we can't talk technology, i.e., Fibre Channel versus iSCSI protocol." But that's not what they want to hear. They're going to want to hear Fibre Channel versus Ethernet. If you go into a small shop with 50 people, there's going to be somebody that knows Ethernet because that's what they're running. They're not running ATM anymore. And you're saying, "We can go in and put in Fibre Channel." Well, if these are the same folks who go out to Circuit City to buy a cheap 10/100 Ethernet NIC and plug it into the new server they just ordered from Dell, or that they bought at Circuit City right along with it, they're going to want to do the same type of thing. These are the same people that are going to want to be able to say. "I bought a new server, I got it from Dell and, gee, it has a Ethernet NIC. I'm going to go out and buy another and tie it into my SAN." In that environment, why would they go to Fibre Channel? They now would have to learn a new technology completely. They have to buy a separate set of switches that are totally foreign to them. So I can't understand why a small to medium business would consider Fibre Channel, because there's no performance advantage.

Wall: It almost seems, from this conversation, that SMB is just one entity. But you could probably slice the SMB market into ten segments. Maybe I'm exaggerating. There's going to be a bunch of folks out there just like you've described. And I don't think Mike Smith of Emulex Corporation would sit here and logically argue that that particular guy should have Fibre Channel. On the other hand, we also know that the markets are price elastic. We know the technology is improving to the point where the cost can be brought down, whether it's through improved process technology or reduction in features or combinations of both. And that the market for Fibre Channel implementations is going to expand at a very rapid pace over the next few years. And at the same time, iSCSI is going to grow as well. But they're going to coexist, and people have to make sure that they have the right amount of information to understand the trade-offs between the two and what's the best for their environment. As Mike pointed out, yeah, if you have two servers and 12 folks maybe it's not the right place for Fibre Channel. He used the example of 15 servers and a lot of people. Fibre Channel is probably going to come into play as an option at a lot lower than a 15-server implementation. So I don't know if it's so much of a debate about who's going to win and who's going to lose. Intel thinks that there's going to be an explosion of growth in both of those types of storage networking implementations over the next five to six years.

Fibre Channel has been successful, continues to be successful. They're doing things to bring down the costs and to deliver it with a much more aggressive price point. And we've all been waiting for iSCSI to take off. We all expected to see accelerated growth two years ago. And a lot of us spent a lot of money so that we were ready for that. And it still hasn't happened. That said, it's beginning to happen. Probably the second half of this year it will happen--they're going to coexist, and they're both going to do well.

Lorenson: I think you've come back to the "shades of gray" issue. Let me just make an argument with regard to this person that you're referring to. Say that there was a Fibre Channel-based solution for this customer (and I'm not advocating this) where putting the HBA into the Dell server he bought at Circuit City, the customer had the same experience as putting his NIC card in there. And that the experience of connecting the Fibre Channel switch looked just like an Ethernet switch was exactly the same. There would be no difference in his experience because even when he deploys iSCSI, he has to have a separate network for it. If you make the equivalent ease of use for Fibre Channel, he may have a choice. It's starting to get there. It's shades of gray, and some of these customers will go to iSCSI right away.

Padovani: Just to be clear, he can buy his Dell server at Circuit City. [laughter]

PARTICIPANTS

Nancy Marrone-Hurley:  senior analyst with the Enterprise Storage Group
                       and roundtable moderator
Anders Lofgren:        vice president of BrightStor product management
                       products at Computer Associates.
Claude Lorenson:       technical product manager for storage
                       technologies at Microsoft.
Mike Wall:             general manager for the storage components
                  division of the net working and storage group at
                       Intel.
Mare Padovani:         senior product marketing manager at Dell.
Mike Smith:            executive vice president of worldwide marketing
                       from Emulex.

Parts 1 and 2 appeared in the June and Aug issues of CTR. To obtain past issues, please contact [e-mail].