FRAMINGHAM (01/24/2000) - When it comes to storage-area networks, sometimes the hype can get a little loud. Network World's Storage Networking Town Meetings are meant to get beyond the exaggerations and find out what's really going on in this important arena. Here we catch up with a recent SAN Town Meeting in Dallas, where the future of the SAN industry was a hot topic.
At the roundtable were Doug Swords of BMC Software Inc., Mark Knittel of Computer Network Technology (CNT), Gary Breder of EMC Corp., Brad Harley of Hewlett-Packard Co., Vicki Vollmar of Hitachi Data Systems Corp., John Camp of Legato Systems Inc. and Dave Hill, storage analyst at consultancy Aberdeen Group Inc. Network World's Executive Editor Doug Barney and Senior Editor Deni Connor led the questioning.
We talk about voice, video and data convergence as being easier to manage and cheaper to buy and install. At the same time there's a bandwidth explosion.
LANs are moving from 10M to 1000M bit/sec. If we're looking to converge data in a world where bandwidth is increasing at such a rate, why are we looking at separate Fibre Channel networks for our storage?
Knittel: Part of the reason is quality of service - there is a very different orientation for building a network for client/ server interaction vs. storage.
Fibre Channel is a predominant technology for storage because it has speed, distance, I/O capability and [numerous] protocol support with a high-speed fabric underneath it. This harmony currently doesn't exist over on the LAN.
How can users plan for the storage explosion? How do they decide what technology to move to, or whether they're going to use network-attached storage or outsourcing?
Swords: You don't make an earthquake decision. You simply view the sheer size of the infrastructure that exists today in most companies and you evolve into it.You may start off with a SAN that consists of two large storage devices that sit in your legacy machine room and are connected by a short piece of fiber cable. There may be a few applications or users accessing that particular environment and it grows over time. You look at what's practical and move to technology that solves a particular problem you have.
Vollmar: Nobody can plan for the future. Folks should keep their options as open as possible. Users should go with building blocks that give them multiple attachmentsto storage, whether it'sa switched fabric, arbitrated loop or direct-attached [storage]. Don't get locked into a proprietary architecture that will limit your options.
How can users determine if the products they buy are interoperable and conform with open standards?
Knittel: You have to look at openness and interoperability on two levels.
Mostly we're talking about the physical connectivity layer - I don't think that's going to be a differentiating factor for long, because standards are coming into play. There's enough activity between the switch vendors and the different forms of connectivity. You'll always be able to interconnect an EMC drive to any type of server. Just like it did in the LAN world, [physical connectivity] is going to go away as a problem. Higher up in the protocol stack with management applications, you are going to have to do a reality check.
You're not going to see much convergence there for a while, because that's how vendors differentiate. You won't, for instance, see EMC supporting a [remote data] connection to a Hitachi disk storage system on the other end any time soon.
Which is better, Fibre Channel or SCSI?
Breder: Both. The trick is, what's most appropriate for what you're trying to accomplish? You need to look at the structure of the bus, how the cache is designed and what the cache algorithms do for you. What's the overall aggregate speed of the bus? What's the aggregate speed into the drive? What's the speed through the host adapters? You need to decide when you put all those factors together, are you getting the functionality you need for today and the future?
Fibre Channel disk drives are appropriate for some and SCSI disk drives are appropriate for others. That answer won't be the same five years from now. But five years ago people said you wouldn't see SCSI disk drives now. And SCSI still predominates.
Knittel: Fibre Channel is better for one reason - it's going to be as cheap as SCSI and its distance is extendible. For example, Fibre can run 32,810 feet between devices, whereas SCSI devices can only be separated by about 40 feet.
Hill: You'll continue to use SCSI if it's directly attached or internal to the server. As you want greater distance or you want flexibility in putting together combinations that you're going to build into the SAN, and as the technology of the storage management software lets you build the SAN more effectively, then you'll start moving more to Fibre Channel.
In a company with 180 servers, 3.5 terabytes of disk, 1.3 terabytes of data and no storage management policy, how can a SAN help?
Breder: You want to consolidate in a way that helps you to simplify how you manage, share and protect the data. Presume that [the data] is scattered all over the place, you've got a lot of people who spend part of their day managing, sharing, protecting all this. If you can optimize your resources through consolidation, it makes sense. If you can accelerate your response to the changes and needs in performance, capacities, sharing and backup requirements for these different applications that make up the 3.5 terabytes, then do it.
How can we convince non-IT executives of the need for a storage infrastructure?
Harley: The impact and features a SAN can provide is more far-reaching than your IT budget. SANs can affect your core business, regardless of what that is.
If you're in e-commerce, SANs should increase your availability, your up-time and the functionality that you can provide to your customers. If you're looking at backup, SANs should improve your uptime and your restore time. Assess what your needs are, what benefit you're providing and you should be able to provide a monetary benefit that's more far-reaching than your IT expenditure.
Does anyone agree with analyst estimates that most online storage will reside at storage service providers by 2005?
Vollmar: There's going to be a growing populace of application service providers that will provide storage services. No one thinks though that outsourcing is going to be the predominant technology for most mission-sensitive, critical data. Storage utilities don't make sense for the general commercial population.
Knittel: I disagree. Over time, storage service providers are going to be a big factor in how storage solutions are deployed. First you'll take a backup copy of your data and put it somewhere else. It's less likely that you're going to take your primary copy of the data and accept the performance delay of having the application processor a long way away from your data. But the more these things increase and the storage points of presence get closer to being ubiquitous, it will be easier and easier to do that.