Ask any system administrator what he or she really wants, and the answer will most likely come down to two things: simplicity and efficiency. The irony, however, is that just about every advance in one of these areas leads to problems in the other. For example, server virtualization has led to inefficiencies on the storage side. This week, I sat down with Brad Winett, senior director of business development at Isilon Systems, to discuss his company’s “next generation NAS” and how this manages both sides of the coin.
Sunshine: What are you seeing are the key issues for your customers?
Brad Winett: One of the things we’re seeing out there is everyone’s talking about storage efficiency. This means a bunch of things, but at the end of day it’s about getting the most bang for your storage buck. In this economy, IT is being really careful about what they buy and making sure they get a fast return on whatever they do choose.
Sunshine: What does Isilon do to address this?
BW: There are opportunities to implement storage efficiently if the technology easily supports it; and there are three things that leap out as being key to making that happen. First, look for the ability to support a truly just-in-time storage model. Pay as you grow. Mr. IT guy should buy only what he needs today and maybe for next month, but no more. If your storage can be upgraded easily — and with Isilon you can truly grow in a matter of minutes — that’s a real benefit. Why pay for disks now when they will sit empty, especially when the cost of those drives have dropped every quarter for as long as anyone can remember? However, make sure that the storage system you first invest in really can scale just-in-time, without the systems administrator needing to worry about lots of volumes, rebalancing load or spreading data around manually. Those are the painful tasks that promote over-provisioning in the first place!
Second, maximize the raw data you can actually place on the disks. We found out our customers on average (verified independently) have data filling 80% of the actual disk capacity they bought from us, which is a lot more than the 25-40% that NetApp’s CEO Dan Warmenhoven recently cited at an Infocomm meeting in Singapore for typical storage utilization in the field. This is because of the native storage efficiency of our architecture. Our customers don’t have to worry about under-allocated volumes and over-allocated volumes; we automatically balance the data across a single large and robust file system that has little overhead associated with it, performs extremely well even when it is filled up, and eliminates any performance hot spots.
The third part is file efficiency. Depending on your biggest pain point, you might want to implement one or another of the data reduction strategies available today. You have Ocarina, which we have as part of our solution for several installations, and others that allow you to store more data in less space, which addresses both CapEx and OpEx — because you’re buying less storage and also reducing the cost of maintaining it. Of course there are other storage tiers that can take advantage of data reduction such as VTL and tape.
Sunshine: When you talk about efficiency, are you also talking about complexity?
BW: Yes. It’s important to look at the whole IT picture. For example, many companies are investing in server virtualization environments to make their servers more efficient. On the storage side, there are all sorts of implications about how those virtual machines are going to access the storage and your data. You need a much more flexible and agile storage environment to support that, especially as the size of the environments and number of VMs gets large.
The core part of our scale-out NAS architecture is that it was built from the ground-up to support transparently scaling a single file system in a single namespace. We take the data and spread it around under the covers so all you know is you can access and read/write the data through any of the nodes. There’s no concept of separate RAID, volume management, and operating system environments; it’s a truly holistic architectural approach that Isilon brings. It is next generation NAS.
Historically, if you look at large NAS installations, the same people who loved the first one hate the tenth one. Each NAS might have 10 file systems, so maybe I have 100 in total. The sysadmins are saying, “… now I have data sitting haphazardly all over the place, but to truly utilize all of my volumes I have to manually spread it around to use my disks efficiently and try to balance the load.” The bottom line of this problem is that it makes it hard to utilize the hardware efficiently. That’s the reason the NetApp’s volumes are only half full. It’s too time consuming to balance so IT just buys more (and more than they really need). We don’t have that problem, because people get to use the whole thing in a very balanced manner. Plus it’s easy to scale. It’s literally one button. One button and automagically the single file system and namespace has grown and is balanced.
Sunshine: Thanks very much for taking the time to speak with me.
BW: No problem. Thanks for your interest!
Brad Winett is senior director of business development for Isilon and is responsible for the company’s partnerships with complementary technology providers. Most recently, he served as vice president of alliances for IBRIX, Inc, a developer of clustered file system software. Prior to IBRIX, Mr. Winett served in various roles at Exavio, DataDirect Networks, Hewlett-Packard, and Hughes Aircraft Company. He received his B.S. degree in electrical engineering from the University of Illinois.
