To SAN or Not To SAN?

That is the Question

By Stan Hume

Beginning in mid-1998, Storage Area Networks (SAN’s) have been the latest trend in the hotly contested computer storage marketplace. Surrounded by much hype, some of it real, some of it not, the challenge for users of computer storage has been to determine which information is real and which isn’t. StorageTek’s I.S. organization strives to mirror the company’s directions and stay on the leading edge of computer storage technology. Therefore SAN’s were a key technology to take a look at.

When first presented with the idea of a SAN, I believed that SAN’s were much hype with some minor substance changes. After much analysis I became a convert. Let me say that I evaluated SAN’s as a customer, not a StorageTek product marketeer.

While it is always fun for technical people to play with new technology, I have a responsibility to solve problems and provide a solid infrastructure for my customers, who, by the way, are all StorageTek employees. My group is responsible for engineering all of the NT and UNIX servers in the Louisville infrastructure. My customers are very particular about several key metrics including system availability, performance and storage space availability. And, of course, they expect all of this at a low cost. They have very low tolerance for anything that affects the key metrics. So, when it came time to evaluate whether SAN’s have a place in the StorageTek environment, I had to evaluate the technology in terms of the key metrics, system availability, performance, and space availability. Lets look at each one individually.

First, system availability is the most critical metric that my customers look at. Prior to our Server Consolidation project, when we had more than 225 servers in a poorly engineered environment, any single server outage was not a major problem since relatively few users would be affected. In today’s consolidated environment several of the servers are absolutely critical. These servers, when down, basically shut down all of Louisville and can affect operations on a worldwide basis. So, what does a SAN do for availability? By providing alternate paths for the CPU to communicate with the disk storage, we now have a fully redundant architecture that compensates for any single failure and in some cases, multiple failures. Another facet of availability that is key is when adding more storage to the environment, system downtime during the upgrade is at a premium. In the past, we had to take a system down, put interface cards in, pull old ones out, add cables, add disk and hope that when the system was booted, it all came up clean, which it rarely did. With a SAN in place, adding disk is as simple as plugging it into the Access Hub, reconfiguring the hub with the GUI and rebooting the host computer – the host computer hardware is never touched! SAN technology, therefore, brings a higher level of system availability, thus helping our first critical metric.

The underlying technology of the SAN has a significant impact on the second key metric, performance. Building a SAN with pure fibre channel-attached storage yields an immediate performance boost from the 40 megabytes per second SCSI transfer rate to the 100 megabytes per second fibre channel transfer rate. While actual transfer rates are somewhat less than the theoretical maximum transfer rates, fibre channel is two and a half times faster, thus helping our second critical metric of performance. The second aspect of performance is in terms of being able to back up the data on the disk devices to tape without sending the data across the Local Area Network (LAN). Until implementation of backups across the SAN, a significant portion of the data moving across our LAN was backup traffic, especially at night and on weekends. Isolating the backup traffic to the SAN removes it from the LAN, greatly improving user response times and our second critical metric of performance.

The third critical metric of storage space availability is also eased by SAN technology. The SAN, in an of itself, doesn’t yield more space (you still have to buy adequate space) but it does allow you to be more efficient in the allocation of space. Prior to SAN technology, adding disk to a system required the addition of interface cards, cables and then the direct attachment of the new disk. The added disk space was only available to the system it was attached to and could not be "carved up" into pieces that were then assigned to different systems. While the software technology to do this is not available as of this writing, it is just around the corner. Veritas, a StorageTek partner, will be releasing its Cluster Server product in late 1999, which will perform this function.

In summary, looking at the three primary metrics that my group is measured on, system availability, performance and storage space availability, SAN’s address all three of them. This directly translates into the business justification for implementing a SAN.

The next question to be answered is "Would we be interested in a SAN if we were not StorageTek?" I can say that the answer to this question is an enthusiastic "Yes!" Anything that helps us improve our customer metrics, reduces complexity and makes it easier to maintain the environment gets my vote every time. The second piece of this is to consider whether the financial payback of a SAN is worth the investment. To understand this angle you need to look at the cost of the SAN. If you make the assumption that when you need additional disk capacity you buy the current technology, which today is fibre channel attached disk, then the incremental cost of the SAN is in the hubs and switches. Depending on the size of the SAN, the incremental cost represents less than a 10% premium, a small price to pay for the flexibility and benefits of the SAN.

The final question to be considered is "If we did not have to use StorageTek products, what would we use to build a SAN?" A SAN really consists of three components, the fibre channel adapter and drivers in the host computer, the SAN hub/switch and the fibre channel disks. While you can buy all three from StorageTek, the fibre channel adapter for the host computer is a low value commodity item that can be purchased from many vendors. The true differentiators in a SAN are in the network devices, the hubs and switches, and the disk. StorageTek is on the leading edge of SAN network devices, especially with the Access Hub product. It combines the best characteristics of a hub with switching capabilities via a very good Graphical User Interface (GUI). The next generation product, the Access Controller, will go even further. These two products are currently the products of choice for implementing SAN’s. Finally, the StorageTek 9145 storage arrays are current state-of-the-art devices, providing fibre channel interfaces from the front end all the way to the disk. While StorageTek OEM’s these products from Clariion, the additional testing and certification that StorageTek does with host computers, network devices and SAN configurations makes the StorageTek-branded devices the products of choice.

In summary, SAN’s are here, they work, they provide value and StorageTek is on the leading edge of delivering SAN solutions.