This week, Chris and Martin are joined by Chris Mellor from Blocks & Files and Eran Brown from Infinidat to discuss flexible pricing strategies in light of current business disruptions. Infinidat has introduced an elastic pricing model for both capex and opex purchases. How is this implemented and what platform features enable Infinidat to deliver this capability?
As businesses are disrupted through the coronavirus pandemic, uncertain times means uncertain demands and budgets. Companies may defer spending or want flexible purchasing models for technology deployments. Public cloud is one choice, but this comes with challenges around data repatriation.
Infinidat uses low-cost media and proprietary data management techniques to optimise I/O across DRAM, disk and flash. With the majority of data on cheap storage, Infinidat is able to over-provision and offer customers much greater flexibility in purchasing options, including capital and operational expense.
Eran takes the team through the challenges, explains some of the issues customers have experienced and then details how Infinidat can offer flexible terms to customers over the lifetime of their relationship.
For more information on Elastic Pricing, check out https://www.infinidat.com/en
Elapsed Time: 00:44:28
Timeline
- 00:00:00 – Intros
- 00:01:10 – Uncertain times, uncertain demand and supply
- 00:02:30 – Budgets are under constraint
- 00:04:30 – The storage industry is in a strong position
- 00:06:00 – Is data centre access a challenge?
- 00:09:00 – Are uncertainties driving businesses to public cloud?
- 00:11:00 – Over-provisioning is one solution to avoid friction of purchase
- 00:13:00 – Moving temporarily to the cloud has business implications
- 00:17:00 – Private and public cloud purchasing are inverted
- 00:19:00 – Flexible pricing allows customers to get best value from budgets
- 00:20:00 – Can all-flash vendors afford to put over-provisioned systems onsite?
- 00:24:00 – What is Infinidat Elastic Pricing?
- 00:29:00 – Deploying to multiple customer sites is a complex process
- 00:32:00 – Tiering to cloud is an impractical solution
- 00:33:50 – How can AIOps and analytics help predict capacity demand?
- 00:37:30 – Minority Report for storage purchases
- 00:38:47 – Welcome to the Storage Bank
- 00:40:00 – Purchasing and consumption models will be more important than speeds & feeds
- 00:43:10 – Wrap Up
Transcript
[bg_collapse view=”button-orange” color=”#72777c” icon=”arrow” expand_text=”Show More” collapse_text=”Show Less” ]ChrisE:
This is Chris Evans recording a Storage Unpacked podcast. And this week I have three returning guests, co-hosts, co-presenters, whatever you want to call them. I’ve got Martin. Hi martin?
MartinG:
Hi Chris.
ChrisE:
I’ve got Chris Mellor.
ChrisM:
Hello folks.
ChrisE:
And I’ve got Eran Brown from INFINIDAT.
EranB:
Hi there.
ChrisE:
How are you doing?
EranB:
Pretty good.
ChrisE:
Excellent.
ChrisM:
We’re doing fine.
ChrisE:
Good. Now, four of us. How often do we get four on a call? That’s always good to have. Lots of different opportunities for conversations. This week we’re going to talk about something we’ve talked about in the past before, which was quite interesting, and it’s that whole idea of new consumption models and how you pay for your technology. And specifically, obviously with Eran here in light of the recent announcements INFINIDAT have made, I think Eran, just before we dive in there, I think this is quite an interesting area because obviously there’s a whole set of challenges around the current market and uncertain times as people like to say, but also I think generally just the impact of the cloud and other things are really making businesses have to really think differently to what they used to do.
EranB:
I agree. And I think we talk about uncertainty as if it’s only one thing. We talk about it in the singular sense, but we have a public health uncertainty around COVID-19. We have financial uncertainty around the recovery of the demand curve, and we have supply and operational perspective uncertainties that may create what we call supply shocks when you want to sell your product and your customers want to buy, but you can’t get your parts out of, for example, China or on the other end demand shocks as a result of some people losing their jobs, that will be change to our demand curves.
ChrisE:
And obviously part of that is just simply that people might be looking at how they’re spending their cash at this particular point, because the uncertainty of not knowing whether their business is going to go down a certain direction or not, whether they’ve still got customers, whether they haven’t paid stuff that they’re not getting income for. Lots of people might be looking at cashflow and thinking that this is a scenario where they have to be a bit more prudent, shall we say?
EranB:
Yeah, and I think it’s all about businesses trying to minimize what they spend now so that they can take a more educated decision a couple of months down the road when maybe things start to clear up a little bit.
ChrisE:
Yeah. Are you seeing any changes Martin in terms of decision-making across the industry? I mean, bearing in mind, you’re token, the end user, if you like.
MartinG:
Oh, obviously everybody’s budget is under constraint at the moment. So people are only spending what they need to spend because people are trying to control their cash flow quite carefully, certainly if your business has been adversely impacted by the current situation. So there’s a lot of thought about that. And I think it’d be interesting as we start to come out a lock down because I think there’s been a lot of decisions which should be made across various industry to defer spend, but we could find it certainly goes the other way. You suddenly find that people who have locked down quite heavily then realize that they want to spend money quickly. They might be coming to end of budget cycles, and then there will be a demand, and then we’ll see…
MartinG:
Eran talks about a supply shock because actually vendors can’t fulfill demand of stuff coming from China, things taking longer to get to places. So I’ve certainly seen a change in buying behaviors and the peers in the industry I talked to, it’s a very similar place at the moment. We started talk about next year at the moment as well. Everybody’s stumped is halfway through ’20. So you start talking about what you’re going to do in ’21. And there may be some refocus about where you’re going to change your investment profiles, what you’re going to invest in. And in the back of everybody’s mind at the moment is second wave anyway.
ChrisE:
That’s a good point. Chris, from your perspective, we talked a few months ago, I think. I guess it was that sort of time, very beginning of this when we first started to see impacts from the virus. We talked about supply chain and we talked about whether we thought we would see supply chain issues from media and stuff coming out of China and so on. Do you still have an opinion as to whether you think that’s a problem at the moment or whether you’ve seen things get worse or better?
ChrisM:
I think things have got better. I don’t think the supply chain problems have been that great at all. And I also think that in the storage industry, demand isn’t going to go down at all. I think businesses that come through this will want to analyze their operations, analyze their transactions and make the best use of the data as possible and that’s going to need more storage. I’m seeing the storage industry actually being in a pretty nice place for this.
ChrisE:
Do you know, that’s something I think I hadn’t thought we would talk about particularly today, but it’d be worth to take a second diversion I think. And you might have a view Eran about this, about whether you’ve seen it, but I think that’s a good point that Chris raised is that businesses might now be challenged to think about how they operate. So some of them might be thinking about things like how they manage working spaces with technology that collects data. They could now be going a lot more online and doing a lot more online stuff. There could well be a lot more data generated as part of this that actually we hadn’t really thought about, and that could be pushing demand a little bit, which is quite interesting.
EranB:
I agree. There’s this joke running around the internet about COVID-19 being the best thing that happened to digital transformation because it forced businesses to take the decisions that they’ve been talking about maybe doing for a couple of years. The joke goes, who drove digital transformation? Was it the CEO, the CIO or COVID-19? And it’s usually COVID-19. And I think, yes, that will mean that we do more transactions online and digitally. Mobile, web, whatever, and that also means that we’re going to be generating more data because digital transactions produce a lot more data to be analyzed, which is great. Now we just have to find a way to make all of these costs efficient so our business can sustain that at a time of financial uncertainty and budget constraints.
ChrisE:
Yeah. There’s some other technical challenges we should consider that too, because clearly one of the things I think nobody wants to do at the moment is head into a data center, and certainly going to be mixing with people or going anywhere where there’s lots of air being blown around and potential for stuff to be spread around without having there being some super efficient filters in place. But Martin, generally from your perspective, we don’t really want to go in the data center that often, but actually at the moment, it might be more tricky to get people who are prepared to go into the data center. So going in and out of that data center, we want to minimize as much as possible.
MartinG:
Oh, you do. But that’s always been the case. It’s been the case for the past 20 years or so. You will walk into a big data center, and there’d only ever be one or two other people in it. You’re very rarely working close proximity unless it’s a big install, and you’re trying to shift heavy kits around. So I’m not convinced the impact is as high as some people like to pretend it is. It’s always been difficult to get into a data center, and so it should be. There’s always been, you’ve always had to go through security, you’ve always set up passes. If you work for a large organization, often there’s change controls to go through. Now before you get anywhere near the data center, you have a temperature check.
ChrisE:
That’s a good point.
MartinG:
Yeah. So I don’t think it’s particularly different. It’s supposed to be clean. Your data center is supposed to be clean.
ChrisM:
Perhaps seeing more of an effect with remote installs or remote and branch offices and internet edge sites where the ability to have the kit delivered and then remotely installed and deployed becomes more important.
ChrisE:
That’s probably true.
EranB:
And I think it could be broadened to say that remote support in general is going to be critical in that time. The more you can do remotely, not just install. How well are you designed for secure remote support to provide proactive support instead of having to send somebody to fix something onsite when it’s already broken down. However, I agree with Martin that if you look at it quantitatively at any given point in the day, there aren’t a lot of people in the data center. But if you think of the number of different people who only come into the data center for half an hour to fix something that broke or to bring some edit the memory for a server or stuff like that, there’s actually quite a high turnout for short periods of time.
EranB:
So it’s still a challenge in the COVID days. We had a customer, I think it was in Spain that had the system go through customs, and then literally the date was already put on the truck. The area around the data center was announced the red zone, and we ended up driving a truck instead of one hour for something like four hours just to drive around the red zone to get into the data center because they needed that extra capacity to drive some digital initiative that was starting to help them deal with COVID-19.
ChrisM:
Martin, has the COVID epidemic at all, encouraged your employer to start thinking about putting more assets in the cloud rather than holding them on premises?
MartinG:
No. [inaudible 00:08:56] at all. I’m hearing this a lot and I can see for SMEs or certainly for smaller companies, but you might want to do so. I think the larger companies, I don’t think it actually particularly impacts them. I want to say no but there are very few cases where we’ve started to think about how we might run high performance remote desktops in the cloud. But apart from that, not particularly. In fact, if you think about what cloud does, cloud changes your cashflow situation. Actually, you lose control of your cashflow when you start putting things into the cloud. You can’t go to Amazon and say, “I want to stop playing for that machine. I’m going to take it out and support because I’ll just turn it off.” Whereas if you’ve got all your stuff on premises, you’ve already… You can make a decision and tell the vendor that you want to change your support levels. You may even feel that you want to take some stuff out of support. You have a lot more control of your cashflow at that point.
ChrisE:
I think that’s a fair point, really Martin and I just want to go back to one thing Eran said a minute ago, the idea of people going in and out the data center doing incremental changes around things like maybe fixing up a server, expanding it or whatever. One of the things in the storage industry that was always hard, perhaps some vendors have done it slightly better now than they used to. And that was the ability to put new capacity in any sort of reasonable level of increment. So when we had fixed rate groups or when we couldn’t be do things as dynamically, when we have to install entire shelves. The ability to go into and both increment the technology and rebalance across that was I think, fairly invasive in a lot of ways. And ultimately going back to what we said earlier, you really want to go into that data center as little as possible.
ChrisE:
And if you can deploy some big chunk of stuff every so often and only call it down when you need it, that’s a more attractive solution than it is to be continually going in and actually saying, “Oh, I’m going to have to put a little bit more in, a little bit more in and then spending all the time rebalancing, but that’s a trade off because if the media and if the technology is expensive, putting stuff in, and that you don’t use, you’re going to pay for one way or another. And that’s a really difficult balance, I think, to get right between the two.
ChrisM:
Unless we’ve got a nice big vendor willing to take up the slack and over-provision the on premises kit, and then let people pay for it as they need it.
ChrisE:
I mean, that’s what people did that Martin. That was a standard model. It did exist. It was there, but that was in the days of disk.
MartinG:
It still happens. It’s a model, which we’re seeing more and more, but proposed that a vendor will put extra capacity on the floor and then as you consume it, because actually what it does is it removes friction from any deal. It means that life is a lot easier and it makes everybody’s life easy. You don’t have to have it installed. If you say I’ve a business demand and you need another half a petabyte and it’s already there, it does speed everything, just removes all that friction. So I think we’re seeing it more. And if you look at most of the costs of a lot of say storage, certainly if you [inaudible 00:11:54] you look up a bomb and you think, “Well, you’re actually paying for the big chunk of that.” Is software and the software is there any way. So you [inaudible 00:12:01] is even easy but hardware costs is actually, if you look at a lot of storage deals, it’s actually not a biggest proportion of 50% of that cost is software.
ChrisE:
Okay. But if you’re the vendor though, that’s a different scenario because the vendors paying for that hardware, even if you’re only paying for the-
MartinG:
Hardware.
ChrisE:
… there’s a fraction of your overall cost of… and there’s still paying that hardware. You’re going to say something I think, Eran there?
EranB:
Yeah. I was talking to a TV channel. I can’t mention their name. I don’t have their permission, but the story they told me was quite interesting. They had a reality show for some extreme sport and they wanted to create some user interaction from marketing perspective. Now this is like week three of the season. And somebody in marketing says, “Hey, we have a lot of viewers who create their own videos. Why don’t we let them upload it? We get free content. The users get a higher level of participation. Everybody’s happy. We get to show ads on top of their videos. That’s great.” They go to IT and it says, “All right. That’s about 15 megs per video times, how many people? Well, we have bought a half a million viewers. We don’t know how many of them will upload the video.”
EranB:
They do the math and they end up needing dozens of terabytes to run the project. IT says, “I don’t have that kind of capacity just lying around. So let me order new equipment and I can have this up and running in three months.” And everybody around the table goes, “What do you mean three months? Three months the secret is already out.” They end up taking this project to the cloud and they pay a lot more as a result but worse than that, they lose the video assets because they don’t have the budget to keep them up and available at the end of the season. So they can’t use them for example, to promote the next season, et cetera. So having that capacity available, whether it’s temporary or longterm basis actually changes how you do business. It’s not just the technical question of, “Can I provision that extra capacity without rebalancing?” It has real business implications.
ChrisE:
I think you’ve just described why the cloud has become so popular. Cloud is not about cost, shouldn’t be cost it’s about the agility to do things like that. To spin things up. Yeah, you potentially lose assets or you don’t quite understand the workflow. In that case, depending which cloud I put it into our end, obviously the central thing is when you finish with it, at least to make sure we don’t lose it is to shift it all into one of the colder storage tiers of your cloud provider and leave it there because actually once it’s actually in that cold storage, you’re not paying an awful lot of money for it. And at least you’ve bunkered it away from a time being.
EranB:
Unless you want to actually retrieve it. You’re right.
ChrisE:
Yeah. But you may want to retrieve it at a later day but at least you haven’t lost it. Because actually now IT, you’ve now got three months to work out what the hell are you going to do with this. “Are we going to keep it in the cloud? What are we going to do?”
EranB:
I think you touched a very good point. Cloud is not about cost savings. Cloud was all about, the promise was elasticity and ease of use and cost reduction. And I think, by the way, for SMBs, that actually is true. SMBs mom-and-pop shops, businesses with 10 different applications, no more. They go to the cloud, they actually save money because they can shut down this one rec computer room that costs way too much to operate. But when you talk about enterprises, when they go to the cloud, it becomes a debate between I can do something on premises cheaply, but usually it will take me longer or I can take it to the cloud where I will pay, I don’t know, a 20% premium on top of what I would pay and I would pay it in OPEX and not in CAPEX, but I can get it down near instantly.
EranB:
So my philosophy is this, in the age of financial uncertainty where you want to keep the cash available in your pocket, you now have a very clear choice. Agility in the cloud or cost efficiency on premises. And of course, as a business, you want to have both. So which one do you have the luxury or the opportunity to change? I’m going to argue that you can’t really have a real impact on the price you pay in the cloud. So the cloud is going to remain agile and expensive. What you can do is look at your private cloud and make it more agile.
ChrisE:
Exactly. So Eran, I mean, that’s the key point, isn’t it? When we look at what vendors are doing in terms of their ability to try and change the purchasing model that they’re offering to their customers. Some of it might be about changing things to being a lease rather than a purchase. Chances are everybody is already doing that. And then you’ve got this whole idea of consumption models and how exactly those are going to work to become more cloud-like. But actually they could be implemented to be even more efficient than the cloud, because as, Martin says once you’ve actually done something in the cloud prices are fairly fixed, you’re paying what you’re paying, it’s advertised, it’s charged, it’s billed, and off.
ChrisE:
Whereas this is an opportunity for your enterprise storage vendors to sit there and think, “Maybe we could do something more creative that works for my customer, and actually helps to deliver to their business.” And some of that is going to have to be around that improved flexibility that allows them to do exactly what you’ve just said.
EranB:
Yeah, and let’s take it back to pre-COVID days, right? If I do something in the cloud, I get approval. I then start the project, I drive the business value and I get the invoice the following month. Whereas on premises, I usually get approval, start ordering things in, racking, stacking, configuring them, redistributing the data, for example, and then I can drive the business value three, four months into the process. So what ideally you would like to have is that cloud-like agility where you can start doing whatever you need immediately while staying with the cost efficiencies of your on premise private cloud. That will be kind of the sweet spot for enterprises when they don’t know where they will need to spend their money.
ChrisM:
Eran, if I were Amazon and a customer said to me, “I’m sorry, I need five petabytes of storage tomorrow.” Amazon could probably deliver it. If I was talking to a supplier who would act in the same way would over-provision my own premises storage so that when I needed to scale up, I could use it instantly. Would they be able to provide the same kind overhead buffer? So if I’d suddenly talked to my IT department as an application unit and said, “I need five petabytes of storage.” I’d say, “Sure, when do you want it?”
EranB:
I would rephrase your question just in one area, and I would say, don’t think of it as a fixed number, like five petabyte, because if you’re a one petabyte customer, you’re never going to need five petabytes. I think about it as a percent. No, but it’s a reasonable question. Think about it as a percent of the amount of data you have. And we have customers with 50 plus petabytes installed and yes, these guys do have five petabytes of accessible capacity to them when they need it on premises. If you think about it as you know, “Can I potentially grow my entire install base in something extreme like 15, 20% then?” Then, yeah. That’s definitely something you can do.
EranB:
Now the question becomes, how much will it cost you to gain that agility? Because if I need to place a lot of hardware on your floor without any commitment for your side as to when you will use it, I need to hedge my bets here. And if I’m hedging my bets using very expensive media or very expensive hardware for that matter, I’m putting out high premium on that. Whereas if I’m doing this with the very cost efficient media and I’m using software innovation to drive the performance, for example, I can actually be a lot more aggressive.
EranB:
And again, I cover Europe, Middle East, Africa and Asia Pacific. I’ve worked less in the US so I can only really comment on those areas of the world, but I think almost all of the systems we ship in those regions have excess capacity because initially the customer will come out with a one petabyte RFP, and we say, “How much do you need today?” And they say, “Half a petabyte.” And we just ask them, “So why are you paying for the extra half petabyte today? We’ll take the risk off your hands.” And they realize they can keep half the cashflow for the project in their pockets, and if five months down the road, it ends up that what they need is not more capacity but actually more compute process, whatever data they have, well that money’s available to them.
ChrisE:
I think that raises an interesting question around how you as a vendor then manage that because you just highlighted one very important fact and that’s that depending on the type of system you’ve got, that technology could either be very expensive or it could be very cheap from a hardware perspective. And hence the reason why when Martin was saying about building materials and hardware cost versus software in that scenario. One of the concerns, I think if you were a business who was trying to build a cloud-like model for selling to your end users, is that you’re going to have to take that capital outlay upfront.
ChrisE:
And if you’re buying say flash, which is obviously significantly more expensive than a disk at this point, you’re going to have to pay for that or at least you’re going to have to account for that on your books, even if the customer hasn’t paid for it. So you’ve now got a very different accounting model. Whereas if you look at say, Amazon. Amazon are benefiting from economies of scale, and the fact that they could have a five, 10 petabytes sitting on the floor that any customer could come and use that day. They’re working on laws of averaging and they can see trends and they can see where the growth’s coming and they can therefore play that curve a lot better because of the shared nature of their environment. Whereas use of vendor has got a bit more of a difficulty there because you have to work out that cost curve or that deployment curve, every single customer you have.
EranB:
You’re right. If you’re taking the risk as a customer on yourself, then you have to do that and you if get it right, you either overspent money or slow down the business initiative. So why not have somebody take that off your hands? So, public cloud is one way of risking that investment by saying, somebody else will do the capacity planning across all my tiers. Another way is to partner with vendors who actually enable that. Now, customers still prefer CAPEX to OPEX, just because of financial reporting reasons, not because it necessarily is better money. However, until today, the trade off was that whenever you do something in capex, you take the risk of planning forward.
EranB:
And a lot of that has to change. Vendors have to enable customers and that’s what we’ve done last month. Vendors have to enable our customers to derisk investment and to use CAPEX, OPEX, or even a combination of both within the same product, as it best serves their business. Because otherwise that vendor is creating friction for that customer, and that customer should be looking for something that solves their problem for this kind of uncertain time.
ChrisM:
Are you seeing customers because they wish to have cloud-like agility, putting pressure or encouragement on vendors to come up with these kind of subscription models so they can better compete with cloud suppliers.
EranB:
So, first of all, there’s a hidden assumption in your question that this can only be done in a subscription model, and that’s not true. We do it with CAPEX, we do it with OPEX, and a combination of both, by the way. And until now, that has only been an OPEX model, which ended up costing customers dozens of percent in premiums. Because effectively the storage vendor becomes your bank. They lease the equipment to you. They become a financing company, and that has to be paid for somehow. And the worst possible scenario for the vendor is of course, that instead of paying dividends or instead of using investor money to develop the company, they end up financing something for their customers.
EranB:
However, if you can do that with a capex, then we kind of now share that cost structure. I can put a minimal premium on that, and we can build a model that actually helps us both. So that’s just to tackle that specific bias, but I think customers are not putting enough pressure on the vendor community. In fact, more often than not, we are the ones pushing the customers to say, “Well, wait a second, if you need a petabyte over three years, but you only need half a petabyte now, why don’t you ask the vendors to show you how they can postpone that expenditure?”
EranB:
It’s still not a common questions that the customers ask. I think the reason goes back to what Chris and Martin talked about earlier, they don’t like people coming into their data centers, racking, stacking, cabling, creating the accidental downtime here and there because they miss cabled something. So I think that’s driving the operational side of the house to say, “Oh, no, just give me everything on day one.” But in the current financial uncertain time, that’s no longer a very good decision to take.
ChrisE:
Okay. So let’s move on then Eran, and let’s talk about what you are actually offering because you’ve hinted a couple of times about… in fact you said last month. Why don’t you take us through what you offer now and what your new product offering is? Because that might answer some of the points that we’ve been discussing already.
EranB:
So we believe that there are multiple forces at play, the CIO who needs to keep the budget available to support the business, the application people who need to get things done very, very quickly. And traditionally that was a trade off. You either kept the money available in your pocket, but then add longer lead times or you spend it in advance and got the project out the door quicker, agility as we call it now. We try to bridge that gap for customers and say, “We will derisk that for you. We will bridge the gap.” If you’re, I don’t know, 500 terabyte, 300 terabyte customer now, and you pay for 300 terabytes today, we will put anywhere from 700 to 2000 terabytes on the ground for you. And we won’t even ask you to commit to using them. And then we will, pre-negotiate the price for every increment, and that increment could be 20, 50, 100 terabytes depending on what the customer wants.
EranB:
And we allow you to consume those as a CAPEX model. That’s what we call the capacity on demand. That’s been around for years and most customers in EMEA and Asia Pacific use that model. However, what we announced last month is that if you’re taking a system with us with a paid base capacity, you can now grow with capacity on demand, which is a CAPEX offering, but you can also do leasing within the same product. So for example if I ship the system to your premise and you have 500 terabytes still unpaid for and unused, you can say, “Can I borrow 50 terabytes to run a large database recovery that I need to do and then shut it down next week?” And the answer is yes, and the price will be, as you said, in the cloud listed, pre-negotiated, everything will be very, very clear.
EranB:
You get approval for your manager, you use the capacity and the following month, you get an invoice. So you get the absolute optimal time to market. You stay on premises. So from a cost perspective, you gain all the benefits of your own premises cloud, and all the operational benefits of using the same backup tools and same user access control management, et cetera. But you gain the flexibility of the cloud and you didn’t have to spend OPEX to do that.
EranB:
Now, if you want to spend OPEX, because this is a project that is at a high risk of failing, or it’s very experimental, or it’s temporary, like a large backup project, sure. We can do that. And it’s all within the same product. So single API, you don’t need to have one OPEX system, one CAPEX system, and then you get silos and you go a decade back in the way IT was managed. You get the best of all worlds in a single solution.
ChrisE:
And you’re achieving this simply because you’re deploying systems that have low cost media deployed in them. And you’re prepared to be flexible, obviously that’s part of the discussion. But inevitably there’s got to be a way to work out how you as a vendor managing to make that CAPEX cost work for you because as you said, you still got to make money. You don’t want to be putting stuff on the ground you’re never going to not make money out off. And you certainly don’t want to be putting equipment in that you then have to invest money in to take out because obviously that’s cost for you. Nevermind anything else to take that out and refurbish it and do something else with it.
EranB:
So here we agree with the other Chris, Chris Mellor that said, I believe storage will keep growing. And yeah, we agree. We don’t really see that often a system that we put on the ground that does not use that capacity at some point. So we take a hundred bets and we assume that two of them will flop, but 98 of them will cover that cost. So we’re okay. The point is not just, how do you drive down the cost? Because if you drove down the cost by just putting some cheap and deep archive storage, now you can’t run your production.
EranB:
It’s about solving the cost challenge while preserving performance, availability, reliability, all the enterprise functionality. So there are many cheap and deep solutions out there, but they cannot really replace an all flash array. It’s about getting a cost efficient solution that can still run your business.
ChrisM:
You’re hinting I think that all flash array vendors could not do this because they don’t have the financial headroom in their own operations to do it.
EranB:
Yeah. Plus the fact that most of them have been losing money even before COVID-19. So where do you get the funds to put more equipment on the ground that essentially you’re financing for your customers?
ChrisE:
I have actually been involved in a couple of projects where we were looking at one with a global US vendor, where we were looking to try and, actually it was a US government agency that we were trying to support. I won’t say which one it was, but effectively they had many, many sites and they wanted an on demand model where at any site they could say, “We want this much capacity.” And there’d be a lead time for deploying it. But at the same time, they wouldn’t guarantee any site would have a certain amount of minimum capacity. And we looked at the cost model and we said, “The capital outlay for us to be able to at least see every single environment would be so huge.”
ChrisE:
It just wasn’t worth betting on the project because we just couldn’t make any money out of it. And that whole idea of how you deploy and how you move stuff around and how you work at what the growth curve is in different customers is a really complex model to get right. And if you’ve got an expensive product, it’s that much more difficult to achieve.
EranB:
I agree. I think businesses will have been looking more and more to their vendors to take that risk off their hands, and if you know, five, 10 years ago, whenever that project happened, that wasn’t very common, I think that’s going to become more and more common as customers realize they have that power.
MartinG:
It’s a sensible move by vendors to do this and you need to obviously have to look at the economics. Because if you’ve got a project which comes along and you require another half a petabyte of storage, whatever it is. If a half petabytes is already on premises, is already there for you to use, you’re going to use it. If it’s not and you’re suddenly thinking, “All right, okay, this is another half a million pounds spend or whatever it is.” You may suddenly find that finance are pushing you into an RFP, an ITT process and the whole thing can go competitive anyway. So that actually is a win-win for the customer’s got the storage, but actually the incumbent vendor actually can actually maintain their stickiness. So perfectly sensible and pragmatic approach to business, really.
EranB:
And by the way, you basically just told the story of the first InfiniBox system that came into production. That’s how we got our first system moved from test and dev environments to production. We had 200 terabytes available and somebody did the very obnoxious thing of walking into the storage teams room and saying, “Oh, by the way, we forgot to open that ticket two weeks ago, but we need 200 terabytes today because the vendor who does the new projects already on site, starting to work on it.” And they had nowhere to place 200 terabytes, and that’s how, what is it? Eight years now ago? The first InfiniBox system came into production because you will always have these oopses in the day to day operations of the IT where you suddenly need something very, very quickly. And again, if you don’t have that, then the only option is to do something in the cloud and pay the premium for that.
ChrisM:
Do you think QLC Flash will make life a little bit easier for all Flash array vendors trying to do the same kind of thing?
EranB:
It’s definitely better than trying to sell SLC or something like that, but again, you still need to pay the full cost of media though. There still a factor between the cost of drives and the cost of flash. And if you can drive performance using software out of a machine that provides the most of the IO from DRAM. And if you think about it, we go, I think we’re at about one to 3% of the capacity is DRAM, and then up to 10%, I think, no 5% is flash in terms of our capacity. But then if you look at the performance profile, it’s 90 plus percent coming out of flash and DRAM and only a small fraction coming from the lowest tier.
EranB:
So I think what’s more likely Chris, is that we’ll see flash vendors increasing tiering to the cloud as a way of cost reduction. And it’s funny to me because when you tier to disk, then an IO will from time to time jump from sub-millisecond to let’s say three, five millisecond response time. When you tear it to the cloud, an IO can jump from sub-millisecond to 70 milliseconds or even more, but let’s say 70 milliseconds. Many applications will not be tolerant for those kinds of latencies. So tiering to the cloud always makes me laugh because it’s exactly what happened more than a decade ago with spinning down the drives, which nobody will use because the applications would crash. It’s the whole thing all over again. So the way to drive down the cost is by the cheapest media available to you, where you run your data.
ChrisM:
Does INFINIDAT, the cost of all the extra media put on its customers sites? Or do you have background deal with your distraint suppliers for them to pay some of the cost as well?
EranB:
The cost of the F the disk media within the system is not that big. It’s not a high portion of the cost. So, that’s what enables us. The fact that the majority of media comes from disks, but the majority of IOs comes from flash and DRAM is what enables that financial model. So we have customers that basically get three, four, five times more capacity than they paid for it, depending on what’s their baseline. And they can grow instantly, and nobody grows you know, three, 400% very, very quickly. So it gives them a very, very long leeway until they actually have to think about, “Well, what’s my next growth trajectory going to be like?”
ChrisM:
But it’s INFINIDAT paying for that extra capacities not that you’re leasing as it were from distraint suppliers behind you.
EranB:
No, no, no. It makes no sense. Disks are so cheap nowadays. That just makes no sense.
ChrisE:
I want to take us in a slightly different direction on this particular the next question, and that’s really to start thinking about you as a vendor and your proactive nature, in terms of how you help your customer work out what to do and how you actually put the right amount of capacity on the floor and so on. We’ve seen over the last four or five years? The rise of a lot more SaaS based solutions that will actually take data from your installation base. It will take data, it will analyze it, it will show you performance problems, it’ll show you all sorts of useful information, but actually this is something we’ve had a discussion about on something else before, whether it was a podcast or something that was written, actually, that’s a really strong selling tool for you.
ChrisE:
Because when you’re looking at how to actually position the amount of capacity on premises for the customer in the first place, when you’re looking at customer growth, and when you know that you might see something in the customer’s environment, that’s going to demand extra capacity. You can be ahead of the game. Now where do you push that to the level of thinking, “Well, we’ll just turn up randomly and put another box in.” And the customer goes, “Why are you here?” You know, whether you would go to that level of putting a box on the floor before they even ask for it might be a step too far.
ChrisE:
But at the end of the day, I think vendors who haven’t got the data from their customers infrastructure, and can’t really see that are going to be almost playing in the dark with these sorts of models, because they just can’t predict the right way to actually deliver it.
EranB:
There’s a funny bit of actually artificial intelligence behind that bait. So when we sent [inaudible 00:35:21] to our cloud analytics, we borrowed an AI philosophy from Facebook, where it turns out that not all customers Chris mentioned before that basically capacity will keep growing, but it doesn’t necessarily grows in a very linear or even predictable way if you just try to say, “What will I have tomorrow?” Because for example, some customers have a lot of temporary dev environments. Consume a lot of capacity, but only for a few weeks. So within our cloud analytics, we actually use a Facebook approach that you have multiple AI algorithms trying to predict what the customer did and did in the past.
EranB:
And then they compete based on the past results, which one scores better. And the algorithm that works specifically best for that specific customer is the one we will present them in the cloud analytics to show them when they will run out of capacity. So we kind of say, “All right, we don’t know which one of these algorithms works best. So let’s take a look at your last year here and see which one would have predicted best based on your actual results, and then project forward into the future.” And right now it’s two algorithms, but we’re always debating whether or not we should add other algorithms into the mix. That really allows a lot of flexibility because again, in the beginning, we don’t need to be right. We have that excess capacity, so we have time to learn. And by the time we learn, we have the model fully trained.
EranB:
We can tell the customer, “Hey, your next box needs to be shipped in by this date, otherwise you might actually run out of capacity.” And you have to remember, these are customers that forget what it is to be running out of capacity. So it’s important to kind of raise that flag once in a while, say, “Hey, there’s a roadblock ahead that you have to make sure you go around.”
ChrisE:
And that’s going to be very specific by data center, by location, by all sorts of different things. The bigger the customer you are and the more complex you are, and more locations you have, the more that model needs to be tweaked to be balanced within different locations, because some locations might just be generating not much data, or that might be the place where a lot of the dev work goes on and therefore it goes up and down. And other areas might be production where they just see smooth curves of growth. So actually that level of data is going to be very, very important, I think. And if you’re a vendor who hasn’t got that level of data, you are going to find, I think a struggle to go back into a customer and actually sell effectively to them.
ChrisE:
And maybe we will get to the day where you knock on the door and you go, “We’ve come to deliver this box.” And they’re going, “Well, we never asked for it.” “But you will in a week’s time.” That’d be a bit scary. Would be like minority report from storage deployment which would be an interesting thing to say.
EranB:
Sure.
ChrisM:
I’m thinking this is very much an enterprise on premises data center approach. You’ve got a big massive applications there. You’ve got a big massive data you can draw on. As you spread outside the central data centers, life is going to get harder for suppliers if they want to do this kind of thing, because the individual installations are going to be a little bit more erratic in how they go forward.
MartinG:
It’s much harder to do at the edge, much, much, much harder to at the edge. Where you don’t necessary have a scale as well.
EranB:
Yeah. We don’t look at the edge as a core business for us. We look at the core, we look at where you can consolidate a lot of data and we had customers consolidating something like 18 frames into a single or two system set up, which makes their lives a lot easier because they read a bubble of the silos, and now you can start automating.
ChrisM:
Well, I think for customers in this situation, it’s great because it’s as if they have a storage bank on the premises and they can just go there and get an overdraft when they need it. And at the same time, the bank ensures they never run out of money, storage capacity in the long term either.
ChrisE:
I think I’d add one other bit to that, and I think that’s to look at where we’re headed in terms of the industry, because as clouds changed our approach to how we consume technology in general, it’s become a bit more flexible and we’re used to that idea of buying and giving back very easily. I think a lot of the advancements we’re going to see in say the storage industry are not necessarily going to be entirely around the technology. I think we’ll see incremental technology changes. Like we know we’ve seen QLC and we’ll see other things like that and so on. But I think in order for vendors to be competitive, they’re going to have to work out how they change the way that they both engage with the customer and actually deliver the products and the services to the customer.
ChrisE:
And if you’re not flexible in those, somebody else is going to come and beat you to it and that’s about the vendor being efficient and tidy and organized, and it’s also about delivering what the customer’s requirements are. So looking wider across the market, and I know we’re talking about your product here in a minute, Eran, we’ll point people in the right direction in a second, but if you’re an end user looking at the market, I think those are the sort of things in the future that are going to be more relevant to you. You’re going to be looking at and saying, “Well, so many of these products are very similar, but what is the underlying architectural design that allows them to deliver a more flexible solution to me from both a business cost, time to delivery type fashion.”
ChrisM:
Yeah. This is another aspect of this I’ve just realized, which is that if a vendor has this kind of relationship with a customer and another vendor comes along and says, I’ve got a faster array or I’ve got a cheaper array, it’s not as simple as that anymore at all, because you’ve got an incredibly sticky relationship with a customer once you’ve got this kind of combined CAPEX and OPEX model that we’ve been talking about.
ChrisE:
Exactly, which is why I think you need to start looking at your vendors in more than just speeds and feeds type aspects. You have to look at it in a wider model and say-
ChrisM:
Oh, makes lives difficult for me.
ChrisE:
No, it makes lives straightforward for you Chris, but it makes it difficult for the customer or end user because then they’ve got now so many other metrics they’ve got to think about and it’s part of this calculation.
MartinG:
I can’t remember the last time seriously, but we actually really talked to a vendor very much about speeds and feeds. When we look at speeds, feeds. One of a very brief discussion, but a lot of it is all about for financial or for commercial relationship. How we transact going forward for speeds and feeds. They’re not very interesting. A lot of the innovation is happening in the transactional relationship.
ChrisM:
Is life more difficult for storage startups now, Martin, [inaudible 00:41:19] enterprises like yours.
MartinG:
Oh yeah. Because for small storage startups, if you were selling to enterprises, I’m not going to tell you how much we pay for storage, but it’s a lot less than lot of people think. So you need to prepare… and I’m sure Eran, is aware of that, but the cost per terabyte for enterprise storage arrays is not high. It’s nowhere near what people think it is versus big enterprises. You’re smaller SMEs or SMBs are getting caned. They’re paying way over the odds for some reasons. This is why if I was an SMB, yeah, I’d move all my stuff to cloud. I’d stand a better chance to get something closer to what the big boys are paying it. But if you go and buy from Dell or NetApp, and you just want to buy a single smallish device, you’re staffed. You’re going to pay a lot of money for that. We get much, much closer to his prices, which you may see. Okay, we don’t pay PCWorld prices for terabyte, but actually it’s not as far off as you may think.
ChrisM:
Yeah. Well, I think maybe at least access kit to SMBs make more money that way, Martin.
MartinG:
I do know funny stories about companies in the past who have turned themselves into resellers as an additional revenue.
EranB:
By the way, what you said right there, Martin is exactly what I said in the beginning. SMBs should definitely go to the cloud. However, they should probably go to their friendly neighborhood cloud that can also provide some consulting services than just going to Azure, AWS, et cetera, where they need the skillset now to manage a cloud. So they replace VMware and Oracle skill shortage with Azure or AWS skill shortage, and that’s just swapping one problem for another, but they should definitely go to a public cloud and stop managing a one rack data center.
ChrisE:
Excellent. Right on that note, Eran, please tell us where people can go to find out more about elastic pricing and what we’ve talked about in this podcast.
EranB:
So by the time we air this, I guess our new website will be up. So just go to www.infinidat.com and on the main page, you’ll see a link to the elastic pricing explanation. By the way, this is not just for primary storage, which is where we’ve been focusing most of the day. This is also for our backup appliance, InfiniGuard and you can also now scale your backups instantly in the same way I just explained about primary storage. I think what you’ll find there is mostly a discussion around IT strategies, and how do you face uncertainty and how do you build a successful IT strategy in light of uncertainty.
ChrisE:
For now, Eran, thanks for joining us. Great conversation and look forward to catching up with you sometime in the future.
EranB:
Thank you very much for having me.
Related Podcasts & Blogs
- #141 – Building Storage Systems of the Future
- #132 – Accelerating Ransomware Recovery with Eran Brown
Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #m3km.
Podcast: Play in new window | Download