Exclusive Interview With SoftIron CEO Phil Straw

Phil Straw, 48, is CEO and co-founder of SoftIron since 2019. Before that, he was CTO of the company he contributed to launch in 2012. Previously he spent 12 years at Heliox Technologies. When he has time outside of SoftIron, he does a lot of water activities, especially diving and is keen on marine technology.

StorageNewsletter.com: As we continue to live in a strange period, how does/did Covid-19 impact SoftIron and your team?
Straw: It seems every company has felt the impact of the Cobid-19 situation. For SoftIron, there are things that have changed, and others that our unique way of operating has helped to minimize the impact. Luckily for us, the way we design, assemble, and manufacture products did not have to change at all – no scratch or dent. We manufacture everything in the U.S. and assemble everything in California, a fact that has allowed us to be somewhat insulated from the impacts of the pandemic, even with things like logistics and regional effects. Interestingly, with all that’s happening, industry is seeing a data migration more to the edge as the realization that human assets, employees, have to be catered to in a more decentralized manner, and thus data centers are starting to migrate in response. 

This creates some secondary problems for them that our offering does very well in servicing, things like the ability to build computing with very low power and very low cooling requirements. For enterprise storage, we have the lowest powered storage products in the world due to our philosophy of building task-specific appliances, as opposed to retrofitting commodity hardware to attack a problem. This approach gives us a great advantage for both performance and TCO, which, in this environment, is what virtually everyone is looking for. 

Additionally, SoftIron has distributed remote offices around the globe, meaning that our operations are set-up to absorb a situation like this. We originally set up this way to be more responsive to our customers – to be able to better meet them on their own terms, on their own turf, and we’re finding out that this sort of ‘distributed company architecture’ has made us much more flexible and resilient for a situation like this. Meanwhile, we just continue to focus on serving the customer with very little disruption. As people, we feel for those individuals impacted by the pandemic. As a company, we’re thankful that we’ve managed to not only minimize the impact it’s had on us, but realize the advantages of how our company is set up to operate and respond to customer needs.

SoftIron is an unique player as you develop specific hardware for open source software. How and why did you start this journey? Are there any specific users behind that idea?
Yes, the SoftIron approach is unique in the industry right now. We’ve engineered a new way of delivering technology that meets the demands of the customer on their turf and in terms that they literally can’t get anywhere else. I’m talking about a full spectrum of benefits – everything from performance and efficiency down to manufacturing security and vendor lock-in. We are innovating in ways that haven’t been explored by the industry because of how difficult it would be for them to reverse course and do the things that we’re accomplishing. It started as a group of technology enthusiasts assembling specific designs for challenging special forces operations environments, where, literally, lives were at stake. Think places like the Mariana Trench. 

People talk about edge computing these days. We’ve built a company asking ourselves not just how we can operate at the edge, but really how we can maximize every benefit at the edge. Everything we were building was custom designed for maximizing reliability, performance, density, and the efficiency of power and heat. This was the genesis of our ‘task-specific’ message, because we were building computer equipment to meet the specific tasks required in remote and incredibly challenging environments. 

Today you see a lot of vendors trying to build from the data center to the edge, and they’re running into all kinds of problems because the commodity hardware that is being deployed wasn’t built for these types of use cases and environments. Building hardware in this capacity, where every aspect needed to be engineered to meet stringent standards – you name it, security, efficiency, performance, reliability, every aspect – we realized the warts that the traditional technology manufacturing paradigm is riddled with, and set out to create something that would address virtually every concern we had as technology integrators. We spent a very long time – 7 years, building technology that is very strategic, and the results are that we have some of the most powerful and efficient data center technology in the world.

SoftIron is a pretty small company, you raised $34 million in Series B recently, what is the plan now?
SoftIron is trying to build a very strategic computer company. What we are probably best known for right now is a really unique productization of open source software defined storage – in particular Ceph, which we consider to be the “Linux of Storage.” To use Ceph as an example, we are offering a productization of Ceph in its fullest form – hardware, software, the manufacturing part, but also the ecosystem and how it interfaces to typical enterprise environments. What is unique, other than the fact that we design our own product rather than assemble it from a computer (so that it’s a dedicated device), is that we try to blend meeting a customer on their own terms and their own turf – at the customer end of whatever solution they’re trying to create. We blend the freedom of open source with the typical enterprise quality you’d see from all the known vendors that have existed in enterprise forever, but are very proprietary and locked in. With a Series B round behind us, we’re obviously financing for growth having proved that we have something very unique, along with a commercial trajectory that is validating our approach and efforts. 

We understand you’re a British company, why did you choose to build your system in California? And when you sell in EMEA, do you still build systems in USA?
One of the things that’s unique about SoftIron is we design, assemble and manufacture all of our products ourselves. SoftIron manufacturing is really part of the design process. Every aspect of what we design and the way it comes together and the way it’s built through the factory is very much part of our product design. It’s part of the DNA of what we’re doing and it’s a very different approach to any computer company I know that exists today. Because we own every aspect of the design, we don’t OEM or ODM. We literally design every component of the electronics ourselves. We build everything from source code, which has a couple of properties. First, we can replicate manufacturing anywhere we want because the premise of how we do things does not necessitate volume in order to get margin – we don’t leverage volume at all as a business premise. That’s fairly unique in the computer business. And because we own every aspect of the design and we don’t OEM or ODM, we’re free to do whatever division of manufacturing we want anywhere in the world. We’ve designed it that way specifically to use manufacturing as a weapon of war in a competitive sense. Our manufacturing knowledge is all internal. We could replicate manufacturing anywhere we wanted in the world in a very short period of time.

What we’ve found, though, is that even as expensive as California and the San Francisco Bay Area is, we can still make products and retain very healthy business metrics as far as margin and cost. So to date, what we’ve done is basically base our manufacturing in California, in Silicon Valley, and have a feedback cycle in terms of product development and everything else. To date we’ve managed to service anyone, anywhere in the world from our manufacturing center in Silicon Valley. Often people do this with their manufacturing center in China, which has way more logistical hurdles in terms of shipping, logistics chains and various other things. In many ways, we’re very flexible with moving product around – in and around the US and Europe is really fairly trivial. Part of what we do in the design and the manufacture actually enables a different model of shipping, which is really interesting. It allows us to move things around in different ordered modules with different ways of packaging of a traditional computer. To date, things are working out really well for us. I’m sure in time, we’ll probably have more than two manufacturing centers.

So for storage, you believe in Ceph, what do you offer exactly?
Storage is where we’ve been focused on, it’s what we’re known for, it’s where we’ve seen traction, it’s where we’ve invested the most in productization. Our HyperDrive storage appliances are essentially built from the source code on up to be the ultimate productization of Ceph. You’ll hear us say that we see Ceph as the “Linux of Storage,” which is the thing we essentially bet the company on five years ago. Ceph has seen a whole bunch of maturity and growth. 

Ceph, being a Swiss Army Knife, does block, object, and file level storage. We’ve productized Ceph to fit into various legacy environments so we can connect to virtually everything in some way inside a traditional enterprise. We find that Ceph is often a fit in places that most people don’t expect it to be a fit. We have success with Ceph. We focus hard on Ceph. We try to get the best productization that we can. We definitely had requests for other things to run on our hardware because the hardware is so exceptional, but to date we’ve never really done anything with that, and don’t really see the need for it. We love Ceph. We love its people. We think the Ceph community has a really good, long-term strategic plan. For storage, Ceph is our focus and it’s working for us. If over time we need to adjust, we have the ability to do that. Presently, our ambition in life is to make the best Ceph productization, the best Ceph ecosystem around it, and also meet enterprises on their own terms, providing an enterprise level experience. If you have the ecosystem, the support, and you have the best product, you’re naturally going to win in that market. That’s what we’re starting to see. So that’s just what we’ve done in storage, and you’ll see us do the same thing in other areas.

For example, in networking, our ambition is to replicate exactly what we’ve done with Ceph and storage. We’re using open source SONiC, which was built by Microsoft and operates their Azure cloud. We introduced our HyperSwitch this spring, which is, in its own right, an amazing productization of SONiC, giving everyone an open source networking solution and enabling the ability to build state-of-the-art scale-out data centers without the compatibility issues and vendor lock-in of proprietary solutions.

Users have plenty of choice on the market and can easily adopt Ceph on their hardware or COTS + Ceph? What is your value proposition and advantages?
The idea of “easily adopting Ceph” has been relative to the knowledge and skill within an organization. This is one of the chief value propositions that SoftIron has with regards to Ceph, in that we have abstracted out virtually all of the complexity of implementing and managing Ceph, making it extremely accessible and intuitive to anyone familiar with storage. It’s no secret that for all of its benefits, working with Ceph can be extremely difficult. There is a considerable learning curve when you are starting from scratch and you need the human resources – the command line warriors – to get things operating at full potential. We’ve eliminated all of this hardship, providing powerful visuals and dashboards that make it very simple to operate one of the most powerful storage architectures in the world. 

But outside of that, we’ve designed hardware that is task-specific to running Ceph. We like to say that we asked Ceph what it wanted, and then built our storage around those desires. And what we’ve built is essentially a storage Ferrari. Due to SoftIron customization, we are able to get performance numbers that don’t exist otherwise in the marketplace where storage is built on commodity computers. And when I say performance, I really mean performance in every aspect. Not only in terms of feeds and speeds, but also in density, in cooling, and thus total cost of ownership. 

Performance is a really big differentiation point for us across the spectrum – we’ll put our performance up against anyone. But aside from base performance, Ceph really shines when it scales, and this is a critical point. We’ve engineered our hardware around the needs of Ceph, and thus, we know how it’s going to operate at extreme scale – extremely well. People don’t run into the surprises with us when they reach scale. There are no latent performance bottlenecks or deficiencies that crop up and force you to react to them. With that, we’re able to offer a peace of mind benefit that you’re not going to run into these types of hurdles as your racks grow. 

The other component that we offer is the “Secure Provenance” aspect which I touched on before. We spent a lot of time architecting our manufacturing process to ensure that our customers can rest assured that they’re receiving a “clear box,” with no hidden extras. This goes all the way down to the ability to audit the source code. Would you be surprised to learn that virtually no other vendors can show you pre-compiled source code that can be audited? SoftIron can. We tell our customers that they don’t have to trust us, they can look for themselves. We’ve built a manufacturing process that offers a level of transparency that is simply not accessible anywhere else in the world. Of course, not everyone finds this important. It’s one of those things where it’s not important until it is. And if it is important to you, you’ll eventually end up in a conversation with us because, right now, there is nowhere else to go – and it’s not even close. It took us a long time to reach this point. We can say with iron certainty that it’s not something that is easy to achieve.

Ceph is an unified storage software, do you sell it more for block, file or object interface needs or mixed of them?
It really varies. There’s a certain personality to what happens after the initial engagement. We have a very fast S3 implementation. We’re very well known for performance. We’ve had a huge amount of success underpinning virtualization – things like VMware, OpenStack, Kubernetes. Typically, in those environments, block storage is the selected. In other environments, we see file storage requirements, which we can handle with CephFS. What tends to happen though is that there are one, or maybe two requirements for access into Ceph. Sometimes there’s legacy attached and we have a product called HyperDrive Storage Router that allows you to attach to things like NFS, SMB, S3, SCSI, various other protocols, that makes it very easy for the enterprise. 

Increasingly, we see that once Ceph goes into a deployment, customers realize that one technology can provide them an incredible amount of leverage. I’ll give you a very typical example that we’ve seen replicated many times. A customer will take a HyperDrive Ceph appliance, and then install a pool of data that they serve virtual machines with. At some point, there is a realization that they need object store and they serve it with S3. Certain applications use S3 and as the data pool grows, they see the performance grow, they see the resilience grow, they see any kind of activity just gets better as you go horizontally and scale with Ceph. That’s just one of the properties of Ceph. There’s a realization inside those enterprises that having one common technology – and something that scales practically to infinity – that as you invest in that one thing, you only have to continue to support that one thing. Because Ceph is a Swiss Army Knife, it tends to become the strategic winner. 

There are so many storage choices that are focused on very extreme performance, or specifically block or specifically object, and they do that one thing very well, but at the end of the day, the enterprise can’t really leverage it past its initial function. So often what we’ve seen is that even solutions that are very focused on doing one thing very well eventually become very marginalized. The reason we like Ceph, and thus what we can do for customers, is serve more and more footprint with one technology. Our customers go away with three things that are all in balance. First, a freedom from vendor lock-in because it’s open source and it’s Ceph. Next, extremely high performance because it’s productized in a task-specific manner as a storage product, not a computer that’s repurposed. The other thing that they get is a nice match with enterprise expectations. A CIO can responsibly say: “Hey I’m going to adopt this open source for the freedom and all of the good benefits, but I don’t have to pay the downside of not having leverage in my organization. Whenever I have to do something slightly niche or turn a corner with the technology or adapt it to something, it’s not going to break.” They’re going to get the support for both the hardware, software and the whole ecosystem. They’re going to get what they expect from the traditional storage vendor, but with the leverage that Ceph gives that they could never expect from a proprietary solution.

Does SoftIron deploy more for primary or secondary storage?
We see a mixture of both. Because of the nature of our productization around Ceph – making a storage product built to maximize the performance of Ceph – we get performance and density numbers that other storage producers can’t get. We’re talking about all areas of the performance spectrum. We can do incredible wire speed performance on even spinning media at multiples of 10GB per node, per one rack unit – we have documents on that. The performance that we are able to achieve, thanks to the way we design and engineer our technology, is truly head-turning. When the lightbulb goes on and people see what we’ve done, they start to consider how it scales, and then they really begin to understand the beauty of our engineering approach. Because of that, we see deployments in places next to supercomputers where HPC is not always a natural assumption for Ceph – at least it hasn’t been; we intend to change that perception. But yes, we find ourselves as primary storage, with customers taking advantage of all the powerful things that Ceph can do, as well as in tier two, more insulated situations. 

We’re also seeing a very interesting thing, which is we see tier one and tier two within one customer. This actually happens often, and what ends up happening is that you have two tiers of Ceph running as two different pools of storage using different technology bases – which has a different cost per terabyte per GB/s. For everything that Ceph is and can do, ultimately what it does is give organizations leverage that didn’t necessarily exist before – and by the way, this always seems to be the case when it comes to quality open source solutions. Today it seems, every CIO has to do twice as much next year with half as much money. There’s always more to do, and how do you actually solve that problem? Very mundane things like “how do I train all of the people in my IT staff to know how to deal with four or five types of storage?” is an issue that we commonly see. That, of course, becomes very hard to do as complexity and scale increases. So even if it’s just matters of training and support, people become really attracted to the leverage model that Ceph enables. 

We’ve seen it happen time and again – people and organizations love open source solutions that give them options that locked-up, proprietary solutions simply can’t offer. We are seeing people shift towards Ceph in the same way that happens with any dominant open source solution. When people realize the power of Ceph and what we’re able to do for their organizations to get the most out of it, it becomes a real eye-opener. 

What is your average deployment in terms of raw capacity? cluster size, etc?
We tend to see deployments as high as 5PB. We’ve seen that multiple times. One petabyte is a typical proof of concept. In Ceph, no one is really buying Ceph to use a few Terabytes. It’s usually at least petabyte intent. But we see 5PB deployments as something that is quite common. 

When users approach SoftIron, is it because they wish to replace a deployed Ceph environment or replace something else like a commercial offering? In that case who are the top 3 commercial vendors you usually replace?
We see a lot of that. Of the things we do that aren’t brand new deployments, often what customers are doing is removing traditional enterprise vendors – I won’t name the names – you know all the typical names. We’ve had success replacing proprietary, enterprise solutions. We’ve seen a lot of customers recognize that SoftIron can provide exactly the same level of service and control over the solution but also bring a level of freedom thanks to its open source nature. If we do our job right, that will never happen, but at least that’s there for the customers as a kind of protection for them. It represents a democratization – freedom for the customer. 

Did Covid change your sales cycle and what is your typical sales cycle?
Not surprisingly, there are some customers, especially in the research community, where we’ve seen Covid actually free up budget more quickly. In other areas we’ve seen some projects slow a little as priorities are re-assessed. Overall though I’d say our course and speed has been affected in a minor way, and typical sales cycles remain similar, ranging from a few weeks for smaller projects to a few months for larger, as well as more structured tender processes.

How do you see OpenStack, Container and Kubernetes adoption within your installed base?
This is an area where Ceph really shines and is gaining momentum. When you do Kubernetes at scale (or a combination of Kubernetes and other virtualization), what ends up happening as you scale – say, 5 or 10PB – is that you start running into certain hurdles. For example, if you’re looking for an open source solution – which most people running Kubernetes are – you inevitably run into the issue of assessing what will service not just the virtualization of Kubernetes through block – but what can actually also move files and do S3 object inside your application space. And so you get to this leverage model that if you want to build a very small Kubernetes cluster, the one-to-one mapping of Ceph and OpenStack, for example, doesn’t exist there.     

There are things that you can do with Kubernetes that don’t necessarily necessitate Ceph, but as you get to scale, Ceph just happens to be the thing that serves Kubernetes best. We’re seeing that people who get to scale inevitably end up looking at Ceph. Whether or not that’s something someone does from the beginning because of an assumption, or whether it’s something that happens at the end because, again, it’s this idea of leverage that you can use one technology, one platform and as it grows it just gets better. We see that with customers a lot. Kubernetes is actually really great for Ceph growth. 

What is your cloud strategy?
In terms of working with cloud vendors? We see many customers now moving to hybrid approaches for security and compliance reasons, to improve performance or just to reduce egress costs. Ceph’s native support for things like S3 makes it ideal to provide the on-prem storage as part of a hybrid strategy. On the other side, we see service providers using Ceph to build cloud and cloud-like offerings because of its ability to scale linearly and its inherent resilience to data loss.

We estimate SoftIron’s installed base between 50 and 100 customers worldwide. Wwhat are their common profiles? Do you sell more in universities, research centers, labs… than corporations? How do you define a typical Ceph and SoftIron buyer?
I think we’re at a very interesting stage in the evolution of the market for Ceph, and software defined storage in general. It’s true early customers tended to be from the open source and Ceph community. Increasingly, though, we’re working with customers that just want really flexible, enterprise storage without the lock-in of a proprietary solution. Our tiny power and cooling footprint is also proving to be really compelling in edge infrastructure type use cases and in markets where power and cooling are proving to be major challenges to growth. Open source SDS, running on appliances optimized to run just that, is really proving to be a winning formula for delivering really good storage.

Any preferred partner and use cases for Ceph deployed on SoftIron appliances?
One of the things about our “Secure Provenance” story – with our hardware being designed and manufactured in USA, and giving our customers the ability to audit everything literally down to the pre-compiled source code, is that we attract a lot of organizations – many of which are the three letter agency types – who aren’t keen on publicizing their installations.

We were recently able to announce that the HyperDrive is being used by the Sandia National Laboratories Vanguard program where we are providing file and object storage for their ARM-based Stria high performance computing cluster. We have installations with one of the world’s largest central banks who have built a large Ceph cluster for self-service HPC analytics. We have a remarkable edge implementation with a major aviation defense organization that is using HyperDrive appliances to capture flight data at the edge for quick processing in disparate data centers around the globe. 

Like many distributed storage software, data protection is built-in with snapshot, replication and erasure coding. Do you see your users add backup software? And why?
It really varies by installation. Ceph, of course, has all of these “enterprise” features that help to protect the data. Depending on how conservative you want to be, you can tune the geometry of your erasure coding with Ceph and provide whatever level of protection you want – so in a sense, it’s tunable. Some organizations insist on doubling their failover capabilities with a backup system. We see a variation. Generally speaking, backup and erasure coding are a great match.

Back on Ceph, how is Ceph driven as RedHat acquired Inktank in the past, RedHat being itself acquired by IBM? Do you see these 2 acquisitions changing the Ceph community?
Maybe and maybe not. I think in many ways Ceph may be more liberated by that acquisition than it was previously. One of the things that’s happened, relatively recently, is the formation of the Ceph Foundation. SoftIron is a founding member of that board. This really gives Ceph a presence beyond RedHat, who is the owner of the Ceph trademark. The project of Ceph is guided by the Ceph Foundation and it is also heavily supported by more than Red Hat at this point. SUSE, also, really focuses on Ceph. Red Hat, now Red Hat and IBM together, have lots of storage solutions which they sell alongside each other. SUSE is a huge contributor to Ceph and focuses on Ceph for a lot of what they do. 

And so the Ceph community is actually quite diverse. If you look at the formation of the Ceph Foundation and its make-up – the activities that are going on within that community, the development effort, it’s probably the most diverse open source storage development in existence. I think this is another reason that it’s an inevitability that Ceph keeps on rolling forward. Ceph, at this point, is largely independent in its future existence. I would say that it’s independent from Red Hat at this point.

How do you sell your products? Any privilege models or you let users pick  the method they prefer? What is your MSRP for 1PB raw?
We do all of it – channel, direct, we try to reach the customer wherever they are. Uniquely, as a small company, one thing that we do – it’s absolutely dogma in the company – is meet the customer on their own terms on their own turf. Often it works better to be direct with the customer, and that’s really the desire. We work through the channel, but we also are usually very high touch with the customer to make sure their requirements are met. We’re more focused on customer satisfaction than transacting shipments. That’s really where we are at as a company. We’ve listened to customers – there’s this adage that you have two ears and one mouth and that ratio exists in nature for a reason. We try to really listen with discipline to customers. That has served us very well in terms of learning how to productize what is important. The bottom line, for us, is that we focus on the customer. So far that’s working pretty well for us. 

Storage is the most conservative sale in the data center. More and more as the leverage model of Ceph and software defined storage plays out, the level of reassurance and the level of risk reduction that the customer wants to see, in the true enterprise sense, is important. People don’t want to buy a computer from somewhere in China, install a CD from another vendor, put them together at scale (which could be millions of dollars and tens of petabytes), and then cross their fingers and see what they get at the end. That’s truly terrifying to people in the enterprise. SoftIron has this great ability with Ceph to say with great determinism and an absolute guarantee, we absolutely know what you’re going to get when you install Ceph on the hardware. It already comes designed and assembled to run Ceph – it comes delivered as a whole package, ready to run at whatever scale you need. You really have this “one throat to choke” model, as opposed to you aggregate everything on site and you install it and see what happens later. The more you get to leverage and the more you get to scale – the more that hoping your hardware can handle scale is just not a fit for a customer. Everything that we do from design, to manufacturing, to productization is all about that: what serves the customer best. All the things I’ve talked about really comes from that. We’ve done all this stuff for a reason, and that generally comes from the customer.

Your question about pricing – again we’ve taken a unique approach in the industry here. We publish our pricing on our website. Seeing the price of individual appliances though doesn’t really get you where you need to be. You need to build a configuration that reflects your use case, so we make available a configurator, answer some questions and you can get your solution price right there.

And finally, what can we expect in the next few quarters from SoftIron?
You’re going to see an expansion of productization. You’re going to see more innovation, particularly on the software side. We will focus on creating solutions as we see data migration, and more things moving to the edge. You will see us focused on low power and cooling and things working in a hybrid cloud environment. In this leverage model of Ceph, I’ll give you an example. Working in a hybrid cloud environment brings the opportunity to do things with Ceph. Running on the edge provides the ability to do that. We see this leveraged model where tier one, tier two, mix of NVMe technology and spinning disk on the other end of the spectrum is integrated across the emerging customer architectures from the core to the edge that we’re now seeing – how do we do that and how do we productize it. 

What we actually built in the beginning, and spent half a decade doing, is building this lego-like framework where we could assemble things from parts in a very modular way. As we go through the next 6-12 months, you’ll see a really wide spectrum of things appropriate almost for the archive market – and things really appropriate for high performance computing with levels of performance that people aren’t used to seeing around Ceph. We’re going to do that on one common hardware platform, which is what we’re already seeing today, the Hyperdrive, but we’re going to expand out the fringes and productize a lot deeper and wider. We should be able to serve in this leverage model, a very extreme base of things with one Ceph productization, basically, and all of the ecosystem that we’ve already built out.

It allows us, as we continue to push the company growth, to serve more and more customers in increasingly diverse deployments, a wider spectrum even within the same customer. We’ve started the shift to other products: transcoding, networking, things like that. We’ll eventually just follow the same pattern that we’re doing with storage with those product lines too: listen to customers, work out where the broader spectrum of product can serve customers in the greatest way. Just keep pushing.

It’s really the tip of the iceberg of what we have planned. I would say that SoftIron has ambitions around being one of the most strategic companies in the world, and there’s a lot more beyond what you already see. All of the technology investment that we’ve made for half a decade is really all about what comes even after the next 12 months. There’s something extremely strategic in there, kind of buried. I won’t go beyond that, but that’s the intent of SoftIron: to do something hellishly ambitious. The next 12 months should seed that for us, and beyond that, we’re really going to show what we mean by strategic intent.