Simplivity attacks the ‘unbearable complexity’ of IT

Hyperconvergence is a relatively new buzzword but Westborough, Mass.,-based Simplivity is already boasting of creating version 3.0 of this emerging IT model. In this installment of the IDG CEO Interview Series, Simplivity CEO Doron Kempel talked with IDG US Media Chief Content Officer John Gallant about how Simplivity’s OmniStack outperforms competitors like Nutanix and claims customers deploying workloads on Simplivity can save 22% to nearly 50% compared to running them on Amazon Web Services. Kempel also talked about Simplivity’s partnerships with Cisco, VMware and Lenovo and explored why it took nearly four years to bring the company’s vision of hyperconvergence to reality.

Why was SimpliVity formed? What was the mission?

If you look at IT infrastructure over the last 25 years since the mainframe we’ve seen what you could refer to as the Big Bang. A lot of technologies emerged and created this very long laundry list of products that you, as an IT professional, need to buy from many vendors and then train many people on and then manage from many panes of glass, so to speak. Some can count up to 10 or 15 products that you need to buy. This has become too complicated.

In 2009 when SimpliVity was founded the simple, or humble, mission was to simplify all of that – basically displace all these point solutions you buy from different vendors into one homogeneous software entity that runs on commodity x86 resources, the same ones that Amazon and Google run on. Deliver the best of both worlds. On one hand, the economics and simplicity of the public cloud.

We offer you IT infrastructure that is as simple to manage and as inexpensive as the public cloud. However, and this is the second world, we will also offer you enterprise Tier 1 capabilities. Sounds impossible, but after 43 months – which is twice as long as your normal IT infrastructure company’s journey from inception to version 1.0 – in April 2013 we launched our product. I can tell you about how that progressed but let me make sure that I address your first question which is what is the mission, what is the problem? The problem is complexity, the unbearable complexity, and the mission is to simplify IT, offering the best of both worlds.

I want to make sure that people understand the different parts of the technology that you roll out. Explain what the OmniStack Data Virtualization platform provides.

In a simple way you can think about it as a data services platform. It offers the functionality, and thus also the replacement, for the following products: One is the storage switch; two is the primary storage array. I can give examples of the vendors that offer these products but I think your readers will understand. We talked about the storage switch and then you have the primary storage array. Then you sometimes have acceleration storage devices such as SSD, flash arrays. Then you have myriad data protection products, sometimes software, sometimes appliances.

For example, you could have a backup application, replication software, DR [disaster recovery] software and that’s a category of products. Then you have another category of products that you can think of as data efficiency devices. Interestingly, since the volumes of data are out of control mainly since the mid-90s, today there are deduplication/compression devices for different phases of the data lifecycle. There are devices that dedupe and/or compress the primary storage. There are devices that dedupe or compress the backup.

For example Diligent and Data Domain focused on that. Then there are devices that focus on the WAN. Riverbed would be a great example. Then there are devices that focus on transit of data to the public cloud.

For instance, Microsoft acquired StorSimple and that is their gateway to the cloud. Then there are ones that focus on archives. There are a bevy of products. Each of them condenses the data at a specific phase of its lifecycle. We have software that offers all these data services and runs them inside a standard server. Basically all the storage services are now assimilated into one software entity that offers a dramatic reduction of cost because all of those functions now run on commodity x86. These are Intel-based servers with great efficiency because you don’t need to overprovision each of them separately because you’re running them all virtualized together and all of that functionality runs on a commodity x86 server.

What is the OmniStack Accelerator Card?

At the heart of complexity is a data problem. In other words, data is out of control. There is too much data to write, to read, to move, to protect, etc. That has given rise to all these appliances that I just described. What we have decided to do is condense, basically dedupe, compress and optimize all the data at inception at the millisecond in which the application writes it for the first time and keep it in that condensed fashion throughout its lifecycle; primary, backup, over the WAN, archive and cloud, globally.

The trick was how do you do that without compromising and introducing latency, reducing performance or [adding] unnecessary complexity? In order to do all of that, we created the OmniStack Accelerator Card. Basically it’s a standard PCIe card on top of which we have a large, very economical processor that has the mission in life of basically condensing the data, applying deduplication, compression and optimization, data efficiency technique.

We do that in real time in near wire-line speed and do that without – very importantly in the world of hyperconvergence – taking away too many of your Intel processors. In theory, everything that we do on the card you could do on Intel processors. In fact, we do that ourselves but if you do that on Intel processors, you take those processors away from the business application. You basically confiscate them in order to run these high-end data services. What we do on the card is compress all the data but we also have a technology that allows us to reduce latency, without getting into too many details, by processing the data very quickly and returning it back to the application. The card is an accelerant. Think about it like a turbo engine in a very fast car. It gives it more speed without taking away from the business application.

I want to ask about the form factors in which you bring this to market. As I see it there’s OmniCube, OmniStack with Cisco UCS, OmniStack with Lenovo. In the OmniCube, that is you delivering this as an appliance with its own x86 server technology?

That is correct. At the end of the day, what the end user gets in all cases is an x86 server with the ratio of compute resources that the customer chose, which is a ratio of CPU to memory, to SSD flash, to HDD. The difference between the two ways we get it to the customer is as follows.

In option number one, we buy the server from the server vendor, we load our software and our card into that server, we put our bezel on it and we ship it to the end user. The end user just receives an appliance with a SimpliVity name.

In option number two, the customer and/or the value-added reseller chooses their own server. Again, they have the same flexibility in terms of the ratio of CPU, memory. Now in order to make things simple for them, they don’t buy the server and buy the software from us. Technically they do but practically what happens is they get delivered to them an already integrated appliance. It’s just that they bought the server. The server that they bought gets to an integration point where we treat it like the mission-critical, atomic IT building block which it is.

We make sure that an integrator that has been trained by us loads into it the software and the card the way that they would have in our own appliance, runs all the tests that ensure that you can run your mission-critical applications on it. This can be 911 applications; it could be hospital operating systems applications and, of course ATM applications. These are real examples.

As end users, what you end up with is an atomic IT building block, which is a server with the ratio that you’ve chosen with our software and our card that was integrated by a professional integration system. In one case it’s a server that we bought for you that is shipped to you with our bezel and in the other it’s a server that you chose yourself but it gets to you integrated with our code anyhow. Those are the two formats. The emphasis here is on the fact that as an end user you have great flexibility in terms of the ratio of compute resources within any one of those systems that you choose to deploy.

Is it possible at some point that it’s a software-only solution that a customer could integrate?

Indeed. Even today, in our engineering organization our engineers run systems that are software only. In the future we will offer systems that are software-only. We’re just going to do it gradually and responsibly because the one thing that we don’t want to jeopardize is the experience of the best of both worlds – a lot of functionality with a lot of resiliency, performance and low latency.

As Intel continues to improve their processing power and to load more Intel cores into the server, we’re getting to the point where running systems that are complete software is a reality without jeopardizing the capabilities that I mentioned and we may release one in the foreseeable future.

As the system becomes more resilient and as everything becomes more mature I can see a reality where the customers eventually load that by themselves. We can do that today. It’s just that we’re a little protective of our customers and for that reason we have the choices of the card, the choice of an integration point before you get the system and also the choice of being very careful in terms of the resiliency of the system.

For example, unlike [other] players in the market, we write data across systems so there is never a situation where the data is only on one system. In addition we have a RAID card in each system, which means that you have a belt and suspenders in order to ensure enterprise resiliency and protection.

In looking on your website it says “the technology is not additive and complementary to in-place data center technology.” I want to make sure that people understand what that means. How do they bring it into their environments if it’s so different from their existing technology?

We would need to look at that statement on the website. I’m not familiar with it. Actually it’s quite opposite. We’ve gone to great lengths in order to make sure that simplicity is also in the way that you purchase and deploy and that we allow you full flexibility to deploy commensurate with, alongside with your existing environment. In fact, we have some very large customers who need to deploy these systems alongside their existing environment.

How does that come about? The way that you move workload, that is applications and data storage systems, is using VMware VMotion, Storage VMotion, these are very basic and brilliant functions that VMware offers that allow you to move an application from your existing legacy environment to our system in a matter that is automated very simply.

Second, and very importantly, we do not force upon you a different management system by SimpliVity that would mean that you now manage us from point of management one, while you manage your legacy environment from point of management two. We philosophically plug our technology, at great effort to us, into existing management technologies and orchestration technologies that are prevalent in the market, for example, VMware’s vCenter or, in the context of our partnership with Cisco, Cisco UCS Director.

It’s a very easy process of basically putting in our systems, connecting them to your network. They immediately self-introduce. They introduce themselves to your VMware environment. They introduce themselves to your management system and you manage them alongside whatever you had in place before that. Some competitors force upon the customer their own management system. That is probably because they want to also manage that particular control point. We want to be very partner friendly to VMware, to Cisco, to anybody else with investment in management. I haven’t met customers who tell me they have a problem with their management technology.

The challenge that they have is with the underlying infrastructure. To summarize, we are completely compliant and complementary with anything that is going on in your environment. What may be meant on the website is that we don’t make your existing servers and storage devices go faster. What we offer is a parallel environment that is very easy to install and manage next to your environment. Sometimes customers ask us if they can run our software on existing servers. Technically yes, but it will introduce such a headache for us and we would require such a large qualification center that today we ask you to not do that. Buy a new server. Let’s make sure that it’s first spec because you’re going to run on it your 911 applications. This is, by the way, a true example. We have cities in the U.S. that run 911 applications on our systems.

Do customers use hyperconverged infrastructure in hybrid cloud? If so, what connections are there with public clouds to create a hybrid cloud?

Yes, customers use hyperconverged infrastructure for the private part of hybrid cloud. Cloud is all about simplifying and streamlining operations and hyperconvergence enhances that on the infrastructure level. With SimpliVity Service Providers, all the benefits of a multi-site private cloud exist in a hybrid cloud implementation. Global unified management, backup policies that exist across sites, easy data migration between sites, and remote backups and restores are all possible between the public and private clouds.

You mentioned that it took 43 months and a fairly sizeable investment — somewhere in the neighborhood of $275 million or $276 million that has been invested in the company — to bring this to market? Why so long? What were the big problems that you had to solve with this product?

When we started this project we asked ourselves: ‘What do we need to do to realize this mission of simplifying IT by virtue of offering cloud economics and enterprise capabilities?’ It was clear to us that we needed to assimilate a lot of functionality that many other startups of the last 15 or so years have developed piece by piece. When we looked at the venture capital industry, what has become clear to us as students of the game, so to speak, is that typically the venture capital companies give you no more than 18-to-24 months of runway to deliver version 1.0 to the market as an entrepreneur because they want to assess the product-to-market alignment.

What you have out there in the market is a lot of 18-to-24-month-deep foundations. Think about it as foundations of skyscrapers and never deeper than 18-to-24 months. The result is that you have point solutions. What we were trying to do is design a new data DNA, a new architecture for the data that is suitable for the 21^st century. Sorry for the bombastic terms. That required us to look at a lot of layers that did not exist before.

Nobody ever deduped, compressed and optimized the data at inception. That was, by the way, one of the concerns that I had with Diligent [Technologies]. {Ed. Note: Kempel was CEO of data dedupe vendor Diligent, which was acquired by IBM in 2008.} I felt that the value that we offered was significant but it was too late in the lifecycle of the data. In addition, managing remote sites required an approach where the data is maintained, deduped, compressed and optimized globally.

Getting to your question, we couldn’t find open-source technologies that would allow us to realize our mission. Had we taken technologies that other startups typically choose – file systems, object storage, etc. – it would have required compromises that would undermine the completeness of the mission. We had to develop our own object store which is a significant piece of technology and, on top of that, our own file system.

In addition we had to develop the accelerator card and then we developed three different technologies: One for accelerated data efficiency that leverages all of that; another one is global unified management which allows us to manage remote sites on a per-virtual machine basis; and third, our own data protection.

For example, about 90% of our customers use our data protection capability and about 65% of our customers turn off any data protection alternative that they had. It took us 43 months just to complete the technology foundation. Think about it as a deep foundation of a skyscraper. But we’ve been developing the technology now for six and a half years and I think it’s becoming clear that the skyscraper, the platform that we are developing, is of a different nature.

I would summarize it as follows: In 2009 there was the emergence of what we call convergence 1.0. These are integrated systems such as VCE, FlexPod [Cisco/NetApp], [IBM] PureFlex – basically integration or pre-integration of existing products. Then in 2011 comes the next generation of convergence. We call that convergence 2.0 and all that is, as powerful and effective as it is, is primary storage in the server on a clustered file system. All the other data services such as protection and data efficiency are not part of those foundations.

Then comes SimpliVity two years later. If you compare us to some of our competitors, we started in September 2009 just like them. They started shipping after 20 months, mid-2011. We continued developing the foundations for two additional years and we think we’ve delivered a much broader set of functions. Today some of our customers, among the largest in the world, are building data centers around SimpliVity, which means that our version of hyperconvergence, call it hyperconvergence 3.0, offers the best of both worlds. It’s simple and economical.

According to the Evaluator Group, if you compare running a certain workload on Amazon Web Services for three years as opposed to running it on-site on SimpliVity, including all the costs; the power, the space, the licensing, everything included, SimpliVity is 22% to 49% less expensive to you, Mr. Customer, than running on AWS. This is counterintuitive. That’s one leg.

In the other world is enterprise resiliency, performance – all the capabilities you expect. This allows us now to go to the largest companies in the world, to participate in the largest RFPs and tell them more than our predecessors in hyperconvergence have done. This is not just a solution for VDI or any other Tier 2, Tier 3 applications. This is not just for your remote office. This is not just for very small companies. The largest companies in the world are now installing our systems and some have already made a decision to implement full data centers on SimpliVity.

That means, John, that if you walk into any one of their data centers you will see our systems. What you will not see, which is transformational, are third-party servers, third-party storage, third-party backup, Data Domain systems, Riverbed. All of that is gone. It’s all running the way you would expect it to run on Google or on Amazon. The only difference is that we offer enterprise capabilities not on very large-scale only but also in very small points of presence. If you’re a dentist’s office, a local bank or a transportation company, you can start small and get the efficiencies, the 3x TCO reduction, with very small form factors that scale all the way up to very large companies.

Let me talk to you about the competitive landscape. The name that we hear a lot as a leader in hyperconvergence is Nutanix. How do you differ from Nutanix?

We have a great respect for Nutanix. We actually wish them the best as they IPO, hopefully soon. The way we think about it is as follows: The story of convergence – or this new trend in IT whereby the Big Bang starts to pull back in – starts with VCE. That’s convergence 1.0 where it gives you somewhat faster time to deployment, reduces some of your management cost but still you’re stuck with a lot of gear that costs a lot of money, requires a lot of space, power and labor.

Then in 2011 comes Nutanix. They are the leader of what we call convergence 2.0, storage in the server and a clustered file system and this is great for VDI Tier 2, Tier 3 applications. They were already selling their product in 2011. We started shipping product in April 2013.

Clearly, VCE has about a four- or five-year lead over SimpliVity in terms of selling but is inferior technologically. Nutanix has a two-year lead in terms of selling over SimpliVity. However, I would argue, inferior technologically. SimpliVity spent more time; therefore we are later to market. However, we have an extremely high success rate in winning business. In more than 80% of the cases where our salesforce or value-added resellers meet a qualified customer – this is a customer that has budget, decision authority and an IT problem that we are suitable for – they end up doing business with SimpliVity.

In my mind, this is now just a game of distribution, just getting enough people competently positioning our product globally. Fifty percent of our sales come globally. We are at a very fast rate adding more resellers and system integrators. We’re winning very large deals. We wish Nutanix a lot of success. It’s a very large market according to IDC. The market consumes $106 billion of the stack that I have described in just 2014. This is a very large market.

However, in my estimation it’s very clear very quickly to customers once we meet with them and once the discussion is no longer about the marketing brochures. We can run a demo where the customer wants a POC {proof of concept} or the customer speaks with a reference customer of ours and it is clear that this is completely different technology, convergence 3.0.

MORE: Nutanix founder’s follow-up act: Cohesity

I want to make sure I pin that down to a couple key points. If you were to take that Nutanix solution, what would you still need to run, for example, that you don’t need to have in a SimpliVity environment?

No. 1 is backup data protection. When you introduce SimpliVity, you do not need backup. Backup is a significant cost center for you. You do not need that with SimpliVity. You need that with Nutanix. The second aspect which is very important is that nobody does data efficiency the way that SimpliVity does. Data efficiency or deduplication/compression is so important that I doubt there is any vendor in the world that doesn’t say somewhere in their marketing collateral that they have deduplication.

But it’s like any car company saying they have some electricity in the car. That doesn’t make them Tesla. It just means that maybe they power the radio with electricity. I would say categorically there is no company that dedupes, compresses and optimizes all data at inception the way we do, which also offers you performance because you deduplicate right. You don’t just reduce the capacity. You don’t just reduce the bandwidth for data that runs over the network. You reduce the total amount of data that is written because IOPS have become one of the most expensive factors in your data centers.

SimpliVity reduces IOPS. Nobody even talks about that today. The whole data management capability is on a completely different cloud with SimpliVity and we eliminate the need for any other data efficiency devices. This is number two.

Number three, which is very powerful, is the ability to manage all of your remote sites. There are very few customers who take IT infrastructure seriously who don’t have at least a DR site because they need to have another point of presence. We allow them to manage everything remotely very efficiently. We call that protected remote offices. It’s a very important capability and I can’t think of large deals that we lose when customers have remote offices.

I’d say these are the three main elements. What it boils down to, it’s all about the data management, all these data services. It boils down to a data virtualization platform. I don’t think that anybody set forth a technological mission to create a new data services platform the way that we have. The bottom line, the customer benefits are a significant reduction in the total cost of ownership and you do not compromise the enterprise capability.

You have a very successful partnership at this point with Cisco. I saw a release that there have been 1,000 systems going out on the UCS [Unified Computing System] platform. Is that correct?

That is correct. We have two very important partners. One is Cisco and the other is VMware. Cisco has been a great partner for us. We pursued them in 2014. We qualified our technology on their servers. What that means, John, is that we allow Cisco customers and Cisco value-added resellers to benefit from the SimpliVity technology and value running on UCS. This has been a remarkably useful relationship for us because with Cisco we entered into very large accounts. Some of them, by the way, are not necessarily prior Cisco customers. We entered together and we have a lot of joint customers right now.

What I’m wondering about there is they also have their own hyperconverged offering, the HyperFlex line. How do your two offerings compare? In the long run, is Cisco going to want to continue your partnership or is this simply assisting them as they get their own hyperconverged business up and running?

I want to make sure that everything that I am saying is interpreted in the very collaborative, respectful spirit in which we view Cisco. I would say the following. First of all, if you think about what we do, this type of technology – not SimpliVity but this whole notion of hyperconvergence – poses an existential risk to companies like HP, EMC, NetApp and Dell because there is almost complete overlap between the portfolio of their products and what we now offer in commodity x86 devices. Cisco has a very large networking business so this would not be overlapping with their networking business.

However, the potential of the software we offer is so great it would be a great way for Cisco to leverage their UCS, ACI [Application Centric Infrastructure] and other technologies and basically now offer storage in the server, a wonderful move for them. I think that what we can learn is that they’ve decided to go into that particular direction.

Indeed, they have an offering that is called HyperFlex that is based on their partnership with a company called SpringPath. Time will tell how that works. I don’t have sufficient evidence in that particular regard. We are already six-and-a-half years into this game and this technology of ours is already in the market for three years and it offers great functionality. I have not heard anybody say that they would offer more functionality or more performance than SimpliVity. It’s just a question for everybody if they can catch up and deliver that which we deliver and time will tell if those alternative solutions by everybody – I’m not referring specifically to Cisco – work.

I would say one more thing about Cisco which is very important. Cisco has traditionally done the right thing by the customers. In other words, they have offered solutions with NetApp and EMC while having their own flash array in Invicta. What we see from them is a continuous collaborative spirit and if the customers and the value-added resellers prefer SimpliVity and prefer SimpliVity on Cisco, I think that that will carry the day. You can think about different permutations or scenarios or outcomes. I see Cisco remaining our very important partner. There are different scenarios in terms of how successful they are with their HyperFlex offering, but we will continue being committed to them and view them as a great partner to offer a solution with on behalf of our customers.

Recently I spoke with Meg Whitman [CEO of HP Enterprise]. Clearly, HP Enterprise is making a big push into hyperconverged infrastructure. What do you think of what you’ve seen from them?

A lot of respect for HP and we would like at some point in time to have the same relationship with HP that we have with Cisco and Lenovo and even Dell. In other words, we allow our end users to run our software on any server. This is the promise that we’ve made to the market. You choose our software; it opens to any hypervisor, any management, any orchestration, any server and eventually any public cloud that you choose.

HP is a great company. I think they have the largest market share in terms of rack-mount servers. We would love to and expect to partner with them at some point and we expect that, like everybody else whom I mentioned, they would need a technology such as the ones that we’ve discussed. I think that they’re developing it. There were announcements about that and we’ll wait to see how that works.

My last question is, is there anything else about SimpliVity that you think is really important for people to know that I haven’t asked about?

We’re growing very fast. I think that hyperconvergence 3.0 opens the game, so to speak. We’re committed to this mission. We have the right resources. We’re responsible and love to offer 3x TCO reduction to anybody out there globally. We have resources around the globe, a result of great investments. We have excellent partners and we’re open for business.