Panasas storage roadmap includes route to software-defined

Panasas is easy to overlook in the scale-out NAS market. The company's products don't carry the name recognition of Dell EMC Isilon, NetApp NAS filers and IBM Spectrum Scale. But CEO Faye Pairman said her team is content to fly below the radar -- for now -- concentrating mostly on high-performance computing, or HPC.

The Panasas storage flagship is the ActiveStor hybrid array with the PanFS parallel file system. The modular architecture scales performance in a linear fashion, as additional capacity is added to the system. "The bigger our solution gets, the faster we go," Pairman said.

Panasas founder Garth Gibson launched the object-based storage architecture in 1999 and served as its chief technology officer. Gibson, a computer science professor at Carnegie Mellon University in Pittsburgh, was a a developer of RAID storage taxonomy.

Panasas has gone through many changes over the past several years, marked by varying degrees of success to broaden into mainstream commercial NAS. That was Pairman's charter when she took over as CEO in 2010. Key executives left in a 2016 management shuffle, and while investors have provided $155 million to Panasas since its inception, the last reported funding was a $52.5 million venture round in 2013.

As a private company, Panasas does not disclose its revenue, but "we don't have the freedom to hemorrhage cash," Pairman said.

We caught up with Pairman recently to discuss Panasas' growth strategy, which could include offering a software-only license option for PanFS. She also addressed how the vendor is moving to make its software portable and why Panasas isn't jumping on the object-storage bandwagon.

Panasas storage initially aimed for the high end of the HPC market. You were hired to increase Panasas' presence in the commercial enterprise space. How have you been executing on that strategy?

Faye Pairman

Faye Pairman: It required looking at our parallel file system and making it more commercially ready, with features added to improve stability and make it more usable and reliable. We've been on that track until very recently.

We have an awesome file system that is very targeted at the midrange commercial HPC market. We sell our product as a fully integrated appliance, so our next major objective -- and we announced some of this already -- is to disaggregate the file system from the hardware. The reason we did that is to take advantage of commodity hardware choices on the market.

Once the file system is what we call 'portable,' meaning you can run it on any hardware, there will be a lot of new opportunity for us. That's what you'll be hearing from us in the next six months.

Would Panasas storage benefit by introducing an object storage platform, even as an archive device?

Pairman: You know, this is a question we've struggled with over the years. Our customers would like us to service the whole market. [Object storage] would be a very different financial profile than the markets we serve. As a small company, right now, it's not a focus for us.

We differentiate in terms of performance and scale. Normally, what you see in scale-out NAS is that the bigger it gets, the more sluggish it tends to be. We have linear scalability, so the bigger our solution gets, the faster we go.

That's critically important to the segments we serve. It's different from object storage, which is all about being simple and the ability to get bigger and bigger. And performance is not a consideration.

Which vendors do you commonly face off with in deals? 

Pairman: Our primary competitor is IBM Spectrum Scale, with a file system and approach that is probably the most similar to our own and a very clear target on commercial HPC. We also run into Isilon, which plays more to commercial -- meaning high reads, high usability features, but [decreased] performance at scale.

And then, at the very high end, we see DataDirect Networks (DDN) with a Lustre file system for all-out performance, but very little consideration for usability and manageability.

The niche is in the niche. We target very specific markets and very specific workloads.

Which industry verticals are prominent users of Panasas storage architecture? Are you a niche within the niche of HPC?

Pairman: The niche is in the niche. We target very specific markets and very specific workloads. We serve all kinds of application environments, where we manage very large numbers of users and very large numbers of files.

Our target markets are manufacturing, which is a real sweet spot, as well as life sciences and media and entertainment. We also have a big practice in oil and gas exploration and all kinds of scientific applications, and even some manufacturing applications within the federal government.

Panasas storage is a hybrid system, and we manage a combination of disk and flash. With every use case, while we specialize in managing very large files, we also have the ability to manage the file size that a company does on flash.

What impact could DDN's acquisition of open source Lustre exert on the scale-out sector, in general, and Panasas in particular?

Pairman: I think it's a potential market-changer and might benefit us, which is why we're keeping a close eye on where Lustre ends up. We don't compete directly with Lustre, which is more at the high end.

Until now, Lustre always sat in pretty neutral hands. It was in a peaceful place with Intel and Seagate, but they both exited the Lustre business, and Lustre ended up in DDN's hands. It remains to be seen what that portends. But there is a long list of vendors that depend on Lustre remaining neutral, and now it's in the hands of the most aggressive competitor in that space.

What happens to Lustre is less relevant to us if it stays the same. If it falters, we think we have an opportunity to move into that space. It's potentially a big shakeup that could benefit vendors like us who build a proprietary file system.