In Conversation: Infinidat CMO Eric Herzog on Strengthening Enterprise Storage

Steve McDowell: I’m Steve McDowell, Principal Analyst and founding partner of NAND Research, and I am here with my old friend Eric Herzog. Eric, is the CMO at Infinidat. How are you doing, Eric?

Eric Herzog: Great. Love to be on with you and congratulations again on NAND research and all the great things you’ve already started to do for the industry.

Thank you. Speaking of doing great things, it’s been a really busy couple of weeks for Infinidat. So last week you made a series of announcements, and we’ll talk about those in a second, but they all group around a common theme, which is all flash performance in a hybrid world. Why don’t you talk about a little bit what Infinidat announced last week?

Well what we announced was three things. First of all, as you know, we started off with a hybrid array, the InfiniBox. And even back then we were already faster than 75 to 80% of the all-flash arrays in market because our unique Neural Cache technology, which uses DRAM. However, over time we realized in addition to that incredibly high performant and highly available hybrid array, we needed an all-flash. Two years ago in 2021 we announced the InfiniBox SSA, All-Flash.

Then a year after that we announced an upgrade. The second generation, the InfiniBox SSA II. One of the key tenets is of course our unmatched low latency, as low as 35 microseconds.

Two years ago we announced our InfiniBox SSA, All-Flash. It allowed us to handle the workloads that the original InfiniBox, which again was faster than 75 to 80% of the all-flash arrays, couldn’t. We brought out two years ago the original SSA All-Flash.

We then updated that a year ago with the InfiniBox SSA II. Larger capacity, cyber resilience technology within InfiniSafe and even faster performance now with latency as low as 35 mics. This launch also all-flash centric. We’re announcing our SSA Express Software which allows a portion of our hybrid InfiniBox, that’s right, our hybrid InfiniBox to behave like a mid-range all-flash array.

We are now doubling the capacity at the upper end of the InfiniBox SSA II. And our entire all-flash family, the InfiniBox SSA is going to get scale up technology. So that allows us to not always ship a fully populated all-flash array, which is what we’ve done historically, but now ship with only 60% capacity or 80% capacity. And then you can buy upgrades as you need them six months later, a year later, two years later, you can upgrade as you need it. That’s the gist of the announcement that we’ve got last Tuesday.

The performance gap between what you’re delivering in all-flash is fairly incredible. What you’re doing with the Neural Cache, that’s some solid engineering that’s a good product. Delve in a little bit into the SSA Express Software. What does that do? Does that narrow the gap even further and give me truly all-flash performance?

What we’ve done is an existing InfiniBox customer, 95% of them by the way, can use the SSA Express Software. It’s a software only play. The software is absolutely free. You don’t have to use it if you don’t want to, but it is part of our InfuzeOS operating system in our 7.3 launch, which is coming out in Q4 as we publicly put in the press release last week.

You then can allocate up to 320 terabytes inside of your hybrid array to function as if it’s an all-flash array and assign applications or workloads to only use the Neural Cache out of DRAM or only use the flash. Now remember why that’s important is the original InfiniBox had three different media layers, DRAM, running Neural Cache, a flash layer, and a hard drive layer. Which means that certain workloads, if it overran both the DRAM and the flash, you actually would be getting performance from a hard drive, which is substantially slower.

With SSA Express, it’s akin to embedding an all-flash array of up to 320 terabytes inside of a hybrid. The main advantage there is today you could use our hybrid array and have four or five workloads that we can’t deliver the right performance. Yet our current all-flash array is way too large. What do you do? You buy one, two or three small all-flash arrays from someone else. The problem there is A; clearly additional CapEx. B; additional OpEx, watts, slots, power, floor space and IT resources to manage those extra arrays.

Now with the SSA Express we’re essentially, it’s as if we’re embedding a 320 terabyte mid-range all-flash array inside of the hybrid and allocating certain applications and workloads to only run with Neural Cache. And if it doesn’t run with Neural Cache absolutely has to come out of flash. The performance is exactly the same as our very large InfiniBox SSA II All-Flash array. You get the same performance but a much smaller capacity to deal with and there are people that only need that small capacity.

Okay. You’re essentially carving up what’s a hybrid array into a virtual all-flash array. Does that take away from the performance of the applications that aren’t tiered to the all-flash?

Not at all. Basically with our Neural Cache algorithms and how we do things, we’re able to give the existing applications that could take advantage of the hybrid they get the performance they need and applications that need faster performance, we guarantee that it’ll come out of the flash. It will get the same performance, 35 mics, as if those applications were sitting in our InfiniBox SSA II, All-Flash array. Yet they’re getting it out of a hybrid.

Let’s switch gears a little bit. Your SSA II is already one of the probably highest capacity storage raise I can buy in its class and you’re doubling the capacity.

Right. So we’re going up now to 3.6 petabytes plus of effective capacity counting data reduction and in the same physical floor space. Let’s take an example.. Let’s say NAND research is a customer and years ago you bought four all-flash arrays of our biggest model. You’re a big company and so you need it. You’ve got tens and tens of workloads running on those floor arrays.

Now you call us up and say, “I’m adding a bunch of new applications that need flash, I need two more.” And we’d say, “Steve, hold it. You may not need to. What size? Oh wait, you can buy one.” So instead of taking two floor tiles, one floor tile, instead of two sets of power and cooling OpEx management, you’ve got them run as well as your IT resources, now you have to do it with only one. So basically you get 50% better costing of all the OpEx parameters, including the IT resources needed to run it.

You get better power, better cooling, you free up a floor tile, all the associated cabling, depending on how you’ve configured your data center that goes into the array, all that cabling goes away and could be used for something else. Obviously one of the big benefits is now you can consolidate even more technology into one floor tile which could save… For example, we have a current customer, they had 27 arrays from the competition, three different competitors. They now run four between the InfiniBox SSA and the InfiniBox, two and two.

Well, if we had this larger model, they would only have one InfiniBox SSA instead of two. But we didn’t have the larger model. There is an example of even more arrays consolidating onto a single array with all the savings in CapEx, OpEx and IT resource management. So very, very strong play from a cutting your cost play.

And lastly, and this is going to sound funny to some people, but obviously economy equals ecology in two ways. First of all, if you’ve got a lower power and cooling budget, you just saved money to the power company. Second thing that’s more ecologic, you are using less carbon cycles. And the third thing is when you do go to recycle, you know this well, whether it be your laptop, your desktop, your servers, your storage, your networking, all those components are filled with toxic chemicals. When you go to retire, if you’re a global Fortune 500 and you’re socially conscious and you need to do the right thing, not to mention certain countries environmental regulation. You’re recycling your computer equipment, including your enterprise storage. Well, because of the toxics it’s expensive and you have more toxics.

If you can consolidate more rays into the InfiniBox SSA II with our larger capacity, you cut down not only what’s going on day to day, but when you get to the retirement phase and you now buy the InfiniBox SSA III, you’re recycling one InfiniBox SSA II instead of two them. Or you’re recycling one of ours instead of five, six or seven of the competition. So A; you’re ecologically more friendly, but B; that kind of recycling is expensive. So now you’re recycling one array instead of six, seven or eight from somebody else saving you money and being better ecologically.

No, that makes a lot of sense and it’s very timely. And you mentioned environmental regulations in the EU. What I’m hearing from IT buyers is we’re starting to write in sustainability metrics into our RFPs and RFQs now globally. Anything you can do to compress the density is fantastic.

I just saw a statistic that one and a half to 2% of global energy use is data center, of that 20% is storage. So again, anything that you can do to collapse that density down is goodness.

You’re doubling the capacity to 6.6 petabytes. That’s a lot of data. What kind of customers are asking for that sort of capacity?

Well, as you know Steve, our focus is, if you will, the global Fortune 2000 and similar entities that are in the governmental sector. And then of course, as you’ve seen for example at the Flash Memory Summit the last two years, two of our hyperscaler customers were the winners at the Flash Memory Summit of the best all-flash array for hyperscaler deployments. We do very well with, if you will, the small and medium cloud providers who are trying to compete against Microsoft, Azure, against AWS and against Google Cloud.

That’s the other segments. Those are our key segments and in that space, whether it be the cloud provider space or these top 2000 enterprises, they need tons and tons of storage. Whether it be hybrid or all-flash. We have one customer, for example, in the telco space, has 127 petabytes of InfiniBox technology, 127. And they’re a global Fortune 500.

It’s very common for us. Our average customer right now, whether they be InfiniBox, InfiniBox SSA or our InfiniGuard purpose-built applies or some combination of the three on average has almost 14 petabytes using us, just us. And as you know, none of the global Fortune 500 use exclusively one vendor. O they’ve got other storage with other vendors as well.

We are dealing with people that need tons and tons of storage. Whatever we can do to cut their capital expenditure, their operational expenditure, and their IT resource allocations is what we can do.

Having a 6.6 petabyte all-flash array gives them that advantage. Not to mention if on the consolidation side, which you just mentioned, the more we can consolidate from the competition, the more they’re saving on both CapEx and OpEx. So it’s an economic play. Obviously the economy is still very rough, as you and I talked about when we saw each other at VMworld. The number one concern in the Fortune 500 survey was economy and the turbulent economy.

Anything we can do, which this whole launch, the SSA Express and the larger capacity and what we mentioned earlier with our scale up, all are designed to not just give you that all-flash performance you need for your applications and workloads, but also do it much more economically and much more ecologically.

NAND Research is a small boutique analyst firm. I don’t need six petabytes of data, but I want to deploy your technology. Do I have to buy a fully populated SSA II?

No. What we’ve added is scale up capability. We didn’t have that before. In the past when you bought the InfiniBox Hybrid or the InfiniBox SSA, you bought a fully populated, every slot was filled with media, in the case of the SSA obviously with flash media. How what we’re doing is for the SSA family, both the new one and the old ones that we have, of course. We are announcing the capability of getting it with 60% populated, physically filled, or 80%, or they still have the option for the 100. If you get the 60% or 80% when you need more capacity, you let us know we sell you an upgrade kit with the additional capacity.

The upgrade is non-disruptive. When you add that new flash to get the capacity larger, you do not turn off your applications or workloads. They continue to run while we’re updating in real time, the storage capacity. You can start with 60 and then six months from now you call up and say, “I need more.” We’ll give you another 20%, we’ll charge you for that. You install it, it’s non-disruptive. Then another year later you say, “Well, now I’m at 80%, that’s not enough, I need another 20.” We can get you up to 100.

That scale up capacity gives the end user and of course our channel partners that work with our end users a lower CapEx entry point to buy because we’re not selling them a fully populated 100% array, they can get it 60%, which means it costs us less on our bill of material, so we charge them less.

Nice.

And then you buy your upgrades as you need. Typical scale up in the storage industry as you need it, you scale up and buy more. And we now offer that on the SSA family, both on the new product and on the original product, both now have this scale up capability.

Excellent. You announced all this a week ago. When’s it available or can I get it now?

The SSA Express Software will be available in our coming InfuzeOS launch, which will be of our software of course InfuzeOS being our operating system. That will be in Q4, middle of the quarter. The scale up capability and the larger capacity SSS II actually has been available for several weeks. We just chose to launch everything all at one time, which makes more sense from your perspective of course as an analyst, and for the end users, and obviously for ourselves.

We’re announcing all together, but you could have bought the double capacity about a month ago and the scale up about a month ago, we’ve just announced it last week. Those are available right this minute and the SSA Express Software, we are talking to customers about it because we do want to do a profile because you may need some additional flash. We have an upgrade kit. The software is free and in many instances you won’t need anything. So literally you’re going to get all-flash for up to 320 terabytes for nothing.

But if you don’t have enough flash in that tier, we will sell you upgrade kits. We’re actively talking to customers now in the installed base of course of what they want and then selling them if they need it, the accessory kit with the additional flash. Not everyone needs it even in the installed base and clearly in the new sale. So if you were calling us up and said, “Look, I’d like to get an InfiniBox in Q4 or Q1. I heard you’ve got some technology that I won’t need to buy a couple of small all-flash arrays because you can put essentially an all-flash array embedded inside.”

Then we’ll talk to those and then when we ship it to them in Q4 or Q1, it’ll be configured the way they want it and they’ll have SSA Express. But in the installed base, we’re talking to people now, but the actual ship date of the SAA Express Software will be mid-calendar Q4, so this coming quarter.

Okay. And good initial reaction from your customers?

Incredible positive reaction, particularly on the SSA Express. The analyst community including yourself has been very positive. Obviously we have a quote from you that we’ve been using in our public materials already. The other analysts have all said, “Very smart move.” And then obviously the SSA II, especially a large capacity gives you a better consolidation play, which ends up in today’s turbulent economy of a CapEx and OpEx play.

And then obviously the partially populated allows you to get our technology for lower price. So people always love getting what they need. And by the way, when you get the 60% or 80%, it’s everything. This is not like Adobe where you get Adobe Acrobat and when you get Acrobat Reader, a bunch of features don’t work. When you get the 60% populated in InfiniSafe, InfuzeOS, Snapshot Replication, FinOps, Neural Cache, all the software technology comes on it and all the guarantees.

Unlike the Adobe model where reader has less features than regular Acrobat, we’re not doing that. What we do is even though we are reducing their entry price, they get all the guarantees a hundred percent availability performance and our cyber resilience guarantees around InfiniSafe they get all the software. We didn’t, if you will, dumb it down, which is like what several software companies do that right there… excuse me, several different software companies where there’s, if you will, a light, a regular and a pro versions of their software.

With our smaller configs, they’re just smaller configs, all the software, all the award-winning, the 65 mics of latency, you get all that. There is no reduction in feature function in the smaller arrays, they’re just smaller. And then obviously with the SSA II larger capacity, we went larger. In this launch on the SSA II family, we’ve gone bigger and we’ve gone smaller, giving more flexibility to the end users and our channel partners that work with them.

I like that model a lot. I appreciate what Infinidat is doing with all batteries included. I’m not sacrificing functionality, you’re not up-charging, no hidden fees and your guarantees you went quickly through them, but some of the best in the industry around data protection and all the others that you talked about.