vali_111 - Fotolia
New Elastifile CEO Erwan Menard said he plans to intensify the startup's focus on scale-out, enterprise-grade file storage for the public cloud, as he tries to fuel the company's growth phase.
The stronger public cloud emphasis will mean changes to the product strategy that Elastifile initially laid out when emerging from stealth in April 2017. For instance, Elastifile designed its distributed file system to run on flash storage. But, Menard said, Elastifile's software will be available with spinning HDDs and SSDs in public clouds, although on-premises deployments will continue to require flash.
Prior to joining Elastifile, Menard was president and COO at object storage vendor Scality. He previously held the same positions at DataDirect Networks, a storage vendor that caters to high-performance computing. Menard also served as vice president and general manager of Hewlett-Packard's communications and media solutions business unit and in various leadership roles at Alcatel-Lucent.
The newly appointed Elastifile CEO recently replaced founder Amir Aharoni, who remains with the startup as chairman. Aharoni was unable to relocate from Israel to the United States, "where we want the growth to be led from," Menard said as part of this Q&A. Elastifile's sales and marketing office is located in Santa Clara, Calif., and its research and development arm is in Herzliya, Israel.
What are your primary areas of focus for the next year and beyond?
Erwan Menard: We were born upon the idea that file storage is here to stay, because a number of workloads in enterprises rely on it, and that file storage should be addressed in a software-defined manner designed for flash. That was the initial DNA of the company, from a product point of view.
Now, if we look at the market, we're observing a growing demand for enterprise-class file storage in the cloud. If you look at the data that's going into public clouds, there's either very cold data for archival or disaster recovery purposes, or there's hot data in very small quantities for workloads that are compute-centric. But there is a huge piece missing, which is all the data residing on NAS in the data center. Why aren't those data and associated workloads in the cloud yet? Because there's no decent enterprise-grade file storage service in public clouds.
At Elastifile, we spent four years developing a modern-age, software-defined file system for flash. And we're taking that intellectual property and focusing on adding a strong, enterprise-grade file system to Amazon and Google and Azure. It's two clicks on Google Launcher, which is their marketplace. We automatically provision a scale-out file system. We definitely aim at doing the same thing on the other clouds if customers choose Azure and Amazon. This is going to happen in the next few months.
Elastifile has a flash requirement with on-premises deployments. Is flash a requirement in the public cloud?
Menard: We designed for flash, because silicon is taking over infrastructure. But you can effectively run it on classic disk. In Google terminology, you can run on so-called PDs, [or] persistent disks, which are groups of SSDs at Google Cloud. Or, you can run it on standard PDs, [or] standard persistent disks, which are effectively classic HDDs. We run on both.
The good thing about designing for flash is that we're able to provide significantly better performance than other solutions out there in the cloud. For example, we are able to provide much better performance than Amazon Elastic File [System] storage. I want to think that's because we designed for the flash era.
Does the ability to run on HDDs extend to on-premises Elastifile deployments?
Menard: No. The on-prem deployment option is to run on bare-metal SSDs.
What significant features are in the Elastifile 2.7 release?
Menard: We are updating the [Google] Launcher experience. That experience is going to be significantly simpler in the way you install. The people who are touching our products in the data center are typically storage admins. In the cloud, sometimes it's an application developer who happens to need storage, or someone who is even less technical. And the first impression people have with the product is extremely important in their decision to adopt it or not.
Also part of the package is what we call CloudConnect. It's a tool that allows you to migrate your data from any NAS in your data center to any cloud. When people are absolutely convinced about the benefits of running stuff in the cloud, they often struggle with moving the data to the cloud. Most of the tools on the market tend to go from one certain type of NAS to one certain type of cloud destination. We've done a tool to go from any to any, and that tool is part of the subscription to our product.
Can users buy CloudConnect as a separate product?
Menard: No. Our goal isn't to become a data-mover company. Our goal is to facilitate adoption of the cloud. The Elastifile software is available as a subscription. And, as part of that, you get CloudConnect.
Can we expect more partnerships, such as Elastifile's OEM deal with Dell EMC signed last year? Do customers want to pick their own hardware and take a do-it-yourself approach, or do they prefer to buy your storage software bundled with hardware?
Menard: I think people want to buy software only, because that unlocks the value chain and allows them to commoditize the hardware and separate software and hardware from a procurement point of view. I think there's a market for software only in the data center -- do it yourself -- that is for sophisticated organizations who decided to continue developing their data center for whatever strategic or regulatory reasons.
That being said, I think the overall trend is effectively slightly different. At the whole market level, the trend is to go to the cloud. The data center is less and less an area where you want to experiment with complicated things. If anything, you want to consume very simple offerings.
So, I think those two trends coexist -- sometimes in the same enterprise account. Frankly, our focus is on the cloud, because this is the next frontier. We're much more involved in conversations around lifting and shifting stuff to the cloud.
Do people want to move everything to the cloud, or do you think the hybrid model will win out?
Menard: I'm not comfortable with the word 'hybrid,' because I'm not sure people are clear on what it means. If hybrid means I have a full stack -- application, infrastructure -- that's delivering a certain business outcome in the data center, and I want to replicate that in the cloud, that scenario does exist.
We have a customer in common with Google, called eSilicon. They are doing chipset design. They've augmented the capacity of their data center on a per-project basis. They don't size for the peak. They size for a lower load. And they run the peak activities in the cloud. They did it with us because they didn't need to modify their application at all when running it in the cloud. That's a bursting scenario. I run peak activities in the cloud and continue running baseline activities in the data center.
Another scenario we see happening is people who are lifting and shifting an entire workload to the cloud. And that creates a period of time where both workloads are in the data center and in the cloud -- the target being to run everything in the cloud. If we want to call that hybrid, then hybrid does exist.
Do you think you may have customers that run your software only in the cloud?
Menard: Absolutely. Four years ago, when we were all focusing on the software-defined data center, we were all undersizing the speed at which workloads could move to the cloud.
Is that why you plan to focus less on OEM partnerships and more on getting your software to work better with more clouds?
Are customers moving their applications to the public cloud? Or, are they just moving their data and leaving the applications running on premises?
Menard: I think the only case where it makes sense to move the data without the application is when you're looking at archiving or disaster recovery. The object stores of public clouds do a great job at that. When you talk about hot data, having an application running in the data center and tapping into a data pool in the cloud may look great on a slide, but I don't think it makes economic sense.
Which vendors do you go up against in competitive scenarios?
Menard: In the cloud right now, the de facto standard -- but it's a fairly low one -- is Amazon EFS [Elastic File System]. Another option is, of course, the status quo: using the same vendor you've been using for decades in the data center and trying to make that work in the cloud. We've seen announcements by the likes of NetApp in that regard. While it's probably a good defensive play, it's very hard with products designed many years ago for the data center to truly take advantage of the cloud. It's going to come with a level of complexity and cost that's probably not viable in the long run.