David Tyler: Panasas has been around for a good number of years now; how would you sum up what the company does and what makes you stand out in the market?
Faye Pairman: Panasas was founded in 1999 by Garth Gibson, co-author of the original Berkeley RAID papers in 1988 and a man generally recognised as a true visionary in the storage space. He was very early to market with a parallel file system, and his vision - since proved correct - was that the world was going to move to a different data type. If you look at the technologies in the '90s and early 2000s, it was all about transactional data - which of course is still very much the case today - but Garth recognised that there would be a growing need for a file system dedicated to the management of really large unstructured data sets by really large numbers of users. This epitomises what we specialise in here at Panasas: moving large numbers of large files, for large numbers of users.

That strength comes from the things that we do particularly well. We scale linearly, which means that as the product gets bigger in terms of capacity, we actually go faster - and that, I believe, is unique. And as the product gets bigger, we actually get more reliable. And the third point - and probably the key differentiator in the space in which we play - is that our system is also supremely manageable. This means that you can deploy Panasas and in a couple of hours you can manage petabytes of data with one or maybe two administrators.

DT: Whereabouts does Panasas sit in terms of its target markets, and the storage sector in general?
FP: Our product isn't typically looking to deploy in a data centre, we're aiming more at the HPC computing sector. Our traditional markets are energy - for instance looking at the ocean floor to figure out the best places to drill for oil, government, with all kinds of applications from nuclear disarmament simulations to systems for building a submarine, and of course all sorts of academic research including material science and life sciences.

Then of course we also work with commercial markets - our product lends itself very well to the commercial space particularly because of its manageability. You could say that manufacturing is our 'sweet spot', along with energy again, pharma, and media/entertainment. So you can see this isn't really high-end HPC, we aim more at the mid-range with that commercial aspect. You might imagine us sitting between a DDN and an Isilon solution, for instance: a lot more large files and general performance than an Isilon, but at the same time a lot more manageable than a DDN solution which might offer 'all-out' performance, but at the cost of a somewhat cumbersome system to manage.

DT: Can you explain in more detail exactly what differentiates Panasas offerings from a user perspective?
FP: Essentially we ship an appliance, fully integrated, so we're responsible for the hardware and the file system. If a user is looking at, say, a DDN solution, they are buying a very sophisticated piece of hardware and then most likely looking to deploy an Open Source file system. So that user needs to be quite confident in their capabilities to both deploy and support that system. The more 'commercial-oriented' approach that we take is that the file system and the hardware are tightly integrated. Why do that? Because it is much easier to deploy and to manage.

It means we can maintain the features and performance of a High Performance solution, crucially combined with that manageability. It gives a HPC solution that scales linearly, and is highly reliable, yet still manageable - that is what makes us unique.

Everything is a trade-off of course, and at the highest end of the federal government for instance, what they want is absolutely no vendor lock-in, extreme performance and low price. And if that system is maybe not quite so reliable, or not quite so manageable, that kind of customer is perfectly capable of working with that solution.

For our typical customers though, they still need that HPC-like performance but typically won't have that level of IT expertise, so our additional manageability aspect becomes essential. Those customers can't get a top-end product like that working if it falls over, they aren't set up to troubleshoot those types of systems that don't have a single vendor behind them. DT: Presumably the channel is key to your marketing?
FP: We have a direct sales force as well as a robust channel programme. We would describe our go-to-market model as 'channel-assist'; the majority of our revenue is via channel partner sales, while the majority of demand is created by ourselves. That's partly because it is a fairly technical sale.

Our channel partners are very important to us in terms of deploying systems, holding government contracts, selling complementary products - so they're crucial to the overall solution sale approach, but in terms of selling the technology, we do a lot of that ourselves.

DT: It sounds as though the Panasas business is evolving from a highly technical focus to one that is responding more to commercial and market demands. How is that reflected in your product development?
FP: I came into Panasas about 8 years ago with a distinct charter, which was to broaden our base. One important aspect of this is an increased disaggregation of hardware and software. As I already mentioned, tight integration is really important for manageability of the product. At the same time we have recognised that there are huge advantages to working with some of the off-the-shelf commodity hardware products that are now available.

With this in mind we recently released our ActiveStor Director appliance, which has been described as like 'the brain behind your storage system': it's a metadata manager, manages global namespace, it manages how you work with the disk management system. We're just releasing that product on a very powerful Intel server, which allows us to take advantage of its processing power and network interconnectivity. This is allowing us to move faster in keeping up with advances in commodity hardware.

DT: So the hardware/software disaggregation focus is based on the flexibility that commodity hardware can offer in the future?
FP: The reasoning behind this disaggregation approach is that we are now seeing great hardware options that simply weren't there in the past, and it allows us greater flexibility in terms of large or small file size performance. File systems are very complex developments, obviously - they're operating systems, in effect - so the move to separating software and hardware will mean a huge uplift in terms of our file system and its next generation. The Activestor Director is just the first release to highlight this approach, and as we progress it will definitely broaden our market appeal - put simply, it will eventually allow us to take our file system and run on more systems than we do currently.

There have been three steps in 'the Panasas process' so far. The early days were all about achieving scalable performance, and the next step during my time here so far has been focused on commercialising the product: taking what was already a really good file system and adding features to make it more reliable and more manageable. The third step is where we are right now, with the move toward disaggregating the software in order to give customers much more choice and a broader product line in the future.
More info: www.panasas.com

"We scale linearly, which means that as the product gets bigger in terms of capacity, we actually go faster - and that, I believe, is unique. And as the product gets bigger, we actually get more reliable. And the third point - and probably the key differentiator in the space in which we play - is that our system is also supremely manageable. This means that you can deploy Panasas and in a couple of hours you can manage petabytes of data with one or maybe two administrators."