We use cookies and other similar technologies (Cookies) to enhance your experience and to provide you with relevant content and ads. By using our website, you are agreeing to the use of Cookies. You can change your settings at any time. Cookie Policy.

Delivering Your Data Faster: A Spotlight Q&A with Gary Orenstein of Fusion-io

Originally published October 11, 2012

This BeyeNETWORK spotlight features Ron Powell's interview with Gary Orenstein, SVP of Products at Fusion-io. Ron and Gary discuss the performance benefits and cost savings that can be achieved through the use of flash memory.
Gary, could you provide our readers with an overview of Fusion-io and your services portfolio?

Gary Orenstein: We're in the business of making the world's data go faster, and that happens through a portfolio of software and hardware products predominantly focused around flash memory to speed up databases, virtualization, cloud computing, big data, and really any enterprise application that needs to perform well in today's world.

Today's world is all about speed. How does Fusion-io accelerate data virtualization and search?

Gary Orenstein: Well, let me take a half step back. It starts with our ioMemory platform and the software to power that, predominantly around our virtual storage layer. The ioMemory products, which come in a variety of sizes and performance points, are called ioDrives. Those are physical products that fit inside of data center servers. They're PCIe-based devices, and the flash memory on those products delivers data much more quickly than conventional disk drives. Customers will use our ioDrives to speed up mission-critical data, placing that data close to the CPU and close to the application within servers so that they can accelerate these applications.

When we jumped into the world of virtual environments, we tackled that with a software solution that pairs with our ioMemory, and that software solution is called ioTurbine. The ioTurbine solution is a caching solution that transforms the ioMemory into a cache so that you can run your virtual machine applications faster, so that you can run more virtual machines per physical server, and so that you can essentially relieve the load and the spending on your external infrastructure. Previously, where people were trying to power virtual environments off of disk spindles, they quickly realized that disks were not able to keep up with the somewhat chaotic nature of virtualized workloads because you have multiple applications on multiple virtual machines accessing the same storage infrastructure. By caching that data on flash memory instead of having to go to disk drives, we can significantly speed it up and provide significant cost savings.

Gary, in many data centers today, performance Ė especially with big data Ė is becoming an issue. How does Fusion-io help data centers eliminate performance bottlenecks?

Gary Orenstein: There's a lot of impact to data center spending that relates to performance and, in some cases, lack thereof. When you're trying to power data from disk spindles, an individual disk in the best-case scenario is going to do something on the order of 100 or 200 I/O operations per second. A flash memory product from Fusion-io will do hundreds of thousands of I/O operations per second, so we're at a totally different scale.

Previously, customers had been trying to configure performance by buying more disk drives, and that can quickly impact the data center because it's more space, power, cooling. Instead of provisioning performance by adding more disk drives, customers can deploy flash memory, place data on flash memory, and basically shrink that whole footprint. But not only are we shrinking the storage footprint, another area of excessive spending in data centers is just buying more servers because the servers and the CPUs are underutilized. By adding flash memory into the equation, we can rebalance the servers so that they're actually being very productive and fully utilized.

There are other impacts too Ė there are reductions in the network infrastructure spending and, of course, reductions in power and cooling. There are a number of areas in data center spending where companies have essentially been running on a partially utilized infrastructure. The CPUs are not fully utilized, and disks are being configured for performance. All of these are no longer necessary with the introduction of data center flash memory solutions.

Gary, when people talked about flash memory, I was always under the impression that it was extremely expensive. It sounds to me that that may not be the case.

Gary Orenstein: I think itís a misconception. Certainly if you look at it from the cost per capacity, one might say it's higher priced, but thatís really not the way to look at it. The way to look at these kinds of architectures and implementations is the cost per transaction. Over and over again, we find that customers are seeing a rapid cost justification and cost savings by thinking about how much it costs them to deliver, for example, a database transaction, or an application transaction. In that light, the flash memory products are actually quite cheap. We have† customers who can deploy databases on one or two Fusion ioDrives instead of having to buy a multi-hundred thousand dollar SAN. A couple of ioDrives, depending on the pricing and so forth might, be 20 or 30 thousand dollars depending on the configuration. A small SAN could easily be ten times that cost.

In the context of solving application problems, flash memory is actually an extremely affordable solution, itís a very efficient solution, and itís very light comparatively on power, and space, and cooling. We have a lot of customers who tell us that the products essentially pay for themselves day one because they avoid having to go out and buy a more expensive conventional storage infrastructure.

Flash memory has been around for a relatively short period of time, and I'm sure our readers want to know what happens if it fails. Is their data protected from loss?

Gary Orenstein: Absolutely. Our ioDrive products themselves are very well protected. Of course, NAND is susceptible to errors and failures; but through a set of sophisticated technologies, weíve been able to protect our customers from any failures. We have a technology called Adaptive Flashback, which mitigates those problems and masks them so that we can keep the product in a state of full health even though they're may be idiosyncrasies with the underlying media. That's part of our Fusion-io software and intellectual property. The devices themselves are very well protected. Then, of course, when architecting for these kinds of solutions, no customer ever wants to have a single point of failure. So we'll frequently work with customers to build out a highly available solution where they might take a database that's on Server A enabled by flash memory and replicate a copy of that database to Server B, which also has flash memory.

In the past, replicating from one server to another server had a performance impact because it was disk drive to disk drive, which as you know just isnít that fast. But now that we can go from flash memory to flash memory, the replication options are much simpler and much easier. For example, we have customers who now will frequently do synchronous replication where every transaction is replicated, and there's negligible performance impact because the flash between the two servers is so fast.

In the past when they had disk-based solutions, they'd have to rely on asynchronous replication where it would be some period of time before the transactions were replicated on that second server. But now it's getting easier to deploy these highly available solutions because of the speed of flash memory.

How easy is it for a company to move applications from their existing infrastructure to Fusion-io? Does it take weeks or even months?

Gary Orenstein: Itís more in the matter of minutes. We have a customer case study video from a company called LockerZ (formerly Plixi). They do photo sharing tied in with social media, and they run a Microsoft SQL Server configuration. Their word is better than mine. In the video, they say it took them about 20 minutes to get it up and running. Not only did they solve basically a whole bunch of performance problems that were happening at the time, but they bought themselves a significant amount of runway for growth without having to change a configuration again. It literally can be as simple as placing an ioDrive inside a server and telling the database to use that ioDrive as the primary data store.

It's somewhat akin to just putting an extra disk drive onto your laptop or USB drive. You just plug it in. The system will recognize it, except in this case with Fusion-io you've now just harnessed the power of a large SAN array from a performance perspective right into a product the size of your hand that can pop right into a data center server.

Can you give us some examples of how your customers are using your solution or the benefits it's providing them?

Gary Orenstein: We have a host of case studies that are available at Fusion-io.com. I'll just give you a couple of examples. Cloudmark, which is a company that focuses on a replication solution, got a 5X improvement when they moved to Fusion-io flash memory, basically replicating technology across multiple data centers. They run a spam mailbox protection solution, and they needed to make sure that all three of the data centers in their geographic area were always on.

We have Callcredit, a credit reference provider that achieved a 25 times performance improvement in their I/O operations for their database. We routinely have customers who will get 10, 20, even sometimes as much as 40X improvement. One of my favorites is Datalogix running their data warehouse queries. These are complicated operations on databases where they're trying to match up a bunch of different variables. Previously Datalogix was running this on a disk infrastructure. When they moved to a Fusion ioMemory, they achieved a 40 times increase in the response time on these data warehouse inquires. It's common for us to hear about situations that would take multiple days for reports to run now being condensed to an hour, or migration plans that were scheduled to take weeks and now take a day because of the performance benefits of flash memory.

Gary, it almost sounds too good to be true. How can a company get started with Fusion-io?

Gary Orenstein: Our products are so affordable that we frequently recommend customers just deploy it with their most performance problematic application. It's not a lot to get started. Our entry-level products are in the $5,000 - $6,000 range. It's not a lot and, again, that's providing the performance that would be equivalent to a SAN that could be $100,000. We find that typically customers will come to Fusion-io because they have a performance problem, but then they'll realize many other opportunities where they can not only solve a performance problem, but also more importantly save money while doing it.

I think the most important thing for customers as they're considering us Ė even if there's not a clear choice from the administrator's viewpoint which applications are problematic Ė is to ask the users if they would like a little more help with a specific application. It's amazing how far the improvements can go and how well received they will be when you take an application that's serving dozens or hundreds of workers in a company and start shaving minutes and hours off their day. Many of our customers report they're making people so much more productive because they're not waiting around.

You mentioned earlier that Fusion-io could be up in minutes. Does it take a rocket scientist to install Fusion-io?

Gary Orenstein: Absolutely not, it is simple. Again, it is just installing the ioDrive, installing the software, and then telling your application to use this drive. When the ioDrives appear in your system, they show up like another drive letter. The advantage of that is that any application that accesses storage can just be assigned to that drive letter.

Gary, this has been very insightful. Thank you for educating our readers about flash memory and introducing them to Fusion-io.

  • Ron PowellRon Powell
    Ron is an independent analyst, consultant and editorial expert with extensive knowledge and experience in business intelligence, big data, analytics and data warehousing. Currently president of Powell Interactive Media, which specializes in consulting and podcast services, he is also Executive Producer of The World Transformed Fast Forward series. In 2004, Ron founded the BeyeNETWORK, which was acquired by Tech Target in 2010.† Prior to the founding of the BeyeNETWORK, Ron was cofounder, publisher and editorial director of DM Review (now Information Management). He maintains an expert channel and blog on the BeyeNETWORK and may be contacted by email at†rpowell@powellinteractivemedia.com.

    More articles and Ron's blog can be found in his BeyeNETWORK expert channel.

Recent articles by Ron Powell



Want to post a comment? Login or become a member today!

Be the first to comment!