Is it better for hosting providers to band together to take on Amazon Web Services or to focus on what each service provider does best?

Ditlev Bredahl OnApp Tony Lucas Flexiant Structure:Europe 2013
photo: Anna Gordon/GigaOM

Session Name: Amazon vs The World.

David Mayer
Ditlev Bredhal
Tony Lucas

Announcer 00:04

Thank you panel. If you’ve been paying attention, and I’m sure you have, you’ve noticed that Amazon is a recurring theme throughout these two days and we’re not going to stop with that now. Up next, we have Amazon versus The World. How can the cloud service providers compete? And that’s going to be a discussion moderated by David Mayer, the Senior Writer of GigOm and he’s going to be talking with Ditlev Bredhal, the CEO of OnApp and Tony Lucas, Founder and SVP of Product at Flexiant. Please welcome our next panel to the stage.

David Mayer 00:43

Ditlev, Tony, thanks very much for joining us. If you don’t know who Flexiant and OnApp are, basically yesterday we had a great discussion of how service providers can take on Amazon in Europe. And now we’re kind of doing a similar topic but we’re stepping back a bit in the value chain, as it were. Maybe not in the value chain but you know what I mean. We’re stepping back. These are the enablers, the people who are going to help these service providers take on Amazon. So let’s just let you now introduce yourselves a little bit and what approach your companies take.

Ditlev Bredhal 01:24

Yeah sure. My name’s Ditlev Bredhal. And I’m the Founder and CEO of OnApp. OnApp’s a software business that helps service providers ultimate their data centers, basically. A part of that is cloud but it’s also CDN, it’s storage, it’s dedicated servers, what’s now coined as smart servers, as well. So we sit as a layer on top of all that stuff and makes it easy for our clients and service providers to deploy it but also for their clients, the inducers to actually manage it. I think that’s pretty much it.

Tony Lucas 01:49

Okay, I’m Tony Lucas, Founder and SVP Product of Flexiant. Flexiant has a similar background in that we’re a software company that enables cloud service providers to build cloud businesses. So actually the way that we look at it is, to help them actually commercialize the cloud, and that’s by enabling them to build that whole suite of services from infrastructures of service and beyond there, how to monetize it, how to build channel capabilities and more. Basically solving the problem but then they’ve got in figuring how to go to market.

David Mayer 02:19

First off, how do you compete against the likes of VMWare and Cloud Stack and Office Back, as well?

Tony Lucas 02:32

I think the easiest thing that both – this is one thing that Ditlev may agree on – Cloud Stack, VMware, all these guys were never built from a service provider focus. There were built from an enterprise focus, and they’ve been tried to turn into – especially VMware – turning into how can we sell this as a service that suits service providers? The technology, the implementation and the billing models for how these things work, none of the technology has ever been designed to solve these kinds of problems, especially the multi-tenant problem. How do you deal with multi-tenanted customers and things like resellers and multi-location and all these things, they’re just not designed to do; a scale especially. But it’s allowing this more [?] they have.

Ditlev Bredhal 03:15

Both Tony and I, both our solutions are basically, built inside hosting businesses. We built our own businesses back in the days. We both sold our businesses and moved on until we have our respective software products now. But it’s true. And also, what they’re doing really well though – the open stacks, the cloud stacks, and VMWare as well – especially lately, it’s leaving this impression of actually ruling this market, especially open stack. If you look at open stack, they have less than 100 and probably less than 50, deployments in the puffy cloud space, where we have more than 2000. Cloud Stack is doing a little bit better. They’re number two in the market after us, with maybe 100, 150 or something like that. But it’s allowing the small market share they have. Yet it’s everyone – you and all the other guys talking about this – it’s all you can focus on. It’s quite interesting how they manage to do that.

Tony Lucas 04:03

You said, “market share.” They’ve got the mind share there and they’ve certainly got the momentum there. But it’s interesting how fickle that is and how much it changes. Before Open Stack, there was another company that was the poster child of open source cloud and they were limited with actions. Same with Open Stack, it’s a great product, in terms of what it’s doing. We’re fairly proud in Stack of the technology, but it doesn’t actually solve the problems for service providers. It does virtual resource orchestration. It lets you spin VMs up easily and do all this stuff. It doesn’t actually help a service provider take the market. There’s no easy way today to go, “Okay, I want Open Stack. How do I go from here and actually have a cloud business where I can actually sell things to customers?” And there’s lots of companies trying to solve that problem but at that point, they’re all proprietary companies. The argument about the open source thing is somewhat different.

Ditlev Bredhal 04:56

And the thing is, if you don’t have something live now, as a service provider, you’re pretty much screwed. You don’t have 6, 12, 18 months that it takes to deploy a free bait, open stack solution.

David Mayer 05:06

What are your customers trying to achieve? Are they trying to tackle Amazon or are they trying to attack niches? Are they going for the vanilla market or what?

Tony Lucas 05:16

It varies very widely and I think it’s interesting that there’s a multitude of problems there. One thing that people are very scared of actually doing is tackling the complete pay-as-you-go market, the pay-per-hour market. And it’s a shame because actually that’s a very powerful market. There’s a lot of potential in there. They don’t understand the revenue models. They don’t understand the margins. They don’t understand how it’s going to cause attrition their existing models. The one thing that I’m glad to see that nobody here is directly trying to attack Amazon, because in terms of the pricing. Here I’m going to do compute per-hour and I’m going to price it cheaper than Amazon. There was a great debate yesterday about, “Can you sell cheaper than Amazon?”, the whole morals, laws and everything else. Yes you can, today. And you may well be able to in the future but ultimately if you try to go into that market, it comes down to one thing and one thing only which is cost of capital. And nobody in the normal set-up market is going to– certainly none of our customers are ever going to be able to out market Amazon in cost of capital. They should do it through the niches is the only way to do it. Find how you differentiate what the key point is, what you specialize in, which may not necessarily be what you think you currently specialize in, and go from there.

Ditlev Bredhal 06:27

I think that’s why these [?] lines that we have, the small ones that are really growing, what they have done, is actually forgetting about cloud and stop building cloud services and stop calling everything cloud and just focus on all the good roles that they have as a business already. But this is actually where I disagree with you a little bit because you said there’s a lot of mileage in the [?] model. We support it as well and we’ve seen some clients deploying it, but I had a question once at a conference I spoke at, where some guy asked me, “Ditlev, how do we convince our long-term ten year clients to stop paying per hour when they have heavily paid per month year after year?”, and that’s the most stupid question in the world. Why would you try to make them change the behavior that actually works for you, works for them, just because it’s cloud? Why should you change the good bill you have with your customers?

Tony Lucas 07:14

No, you’re absolutely right. I certainly wasn’t suggesting they should change their behavior because [inaudible] afraid of is the customers trying to force them into changing the service provider’s behavior. We have class examples but years ago when we talked to primarily football clubs and things like that, you can imagine their traffic spikes are massive and they said, “That’s all right. We like all these capabilities. We love the fact that we can scale up and scale down and everything else, but we want to fixed price every month, because that makes us comfortable.” I said the only way we can do that is to over-provision what you buy so that we can guarantee that we’ll [exume?]. Well, that’s fine. We’ll just deal with that; that’s fine. There’s several different revenue models that can be found in the cloud model kit. There’s a typical VPS model. There’s what we call, resource pools, where you buy a certain amount, you get agility behind it. There’s private cloud and hybrid cloud type of stuff into the way it’s segregated and then you go all the way to the pay-as-you-go. And I think that’s the point, which is all these models are valid, depending on the existing customer base.

Ditlev Bredhal 08:17

I think my point was, it’s not about the cloud. The cloud is only 10% of the market. It’s a 66 billion dollar market we are all fighting for or all clients are fighting for and it’s only around 6 billion that’s actually cloudy. I think that we have to focus on the workloads. So where workloads are best suited. Some might be best suited for dedicated servers, some for virtual servers, some for smart servers, some for even chat hosting in some cases. Is a CDN needed? What kind of storage solutions are best and so on and so on, rather than get washed into this cloud discussion. I think in a few years -hopefully, and it would help all of us, if they did – if they just forget about the term ‘cloud’ and just realize that this is how stuff is done. I think the important part is to have one interface to it. So one UI, one pane of glass in front of all your infrastructures. So, the thing about workloads, they don’t care about the cloud. They just want to work. And if you force them on the cloud just because it’s a new fad these days, you’re missing out. There’s like this rec space versus software. And Pier One, one of our clients, shares software with you as well. It’s all about providing the volt power where you need it.

David Mayer 09:16

How high should these service providers be aiming, in terms of scale?

Tony Lucas 09:22

It’s a good question. There’s two things. There’s scale and there’s what services they’re actually offering. Because one of the biggest things we’ve seen with certain service providers that have come up and said, “Oh, we’re going to do Cloud now and we’re going to do this and we’re going to do that.”, it’s basically rebadge for what they’ve already done. It might be reasonably clever, but it’s still infrastructure. They’ve got to be looking beyond infrastructure in terms of what they’re trying to sell, because ultimately more and more people are moving workloads out to the cloud, and what they’re actually doing – yes I think we’re all fed up with the word cloud – what they’re actually doing is looking at moving business apps out, whereas service providers or hosting providers are far too used to, “Well, a customer wants to host a website.” I remember talking to one hosting company who’s been in the industry for years and years, and he says to me, “All my customers care about is if their email gets delivered quickly and the website loads fast.” There’s so many other use cases out there and other value-added services that need to be done on top. In terms of the actual scale thing, it’s smaller than most people realize, I think, in terms of what becomes viable scale. One rack is not viable scale, but neither is needing to build your own physical data centers either. There is a happy ground in the middle.

Ditlev Bredhal 10:35

My old business, I ran a company called UK2Group, prior to OnApp. We hosted a large retail chain. We hosted them for quite some years, and every second year we had to re-win them as client. I lost them, because they came and they asked me, “What would the terms be if we wanted 1,000 servers from you?” Being bootstrapped, we were on $50 million revenue, good business, but bootstrapped, debt free, didn’t lease anything, I had to ask for at least a year contract term, otherwise I couldn’t take the risk of buying 1,000 servers. I lost a client, because they went to Amazon and picked up 1,000 servers like this, no contract, could give them back in principle if they wanted to in the next hour. I think that’s where scale comes in, and I guess that’s back to the title of why we’re sitting here, right, how do we deal with the Amazon kind of scale, and should we try to? From our perspective, I think the only way to deal with it is– how many service providers do we have in here? Let’s say 30 service providers in here. Let’s say you have a hard time dealing with Amazon. Let’s say you guys sat in a room together and decided, You know what, together, we probably have 30% across our data centers that is spare capacity. A lot of that would already be powered up. Let’s say you guys decided to pool that capacity into one large pool of resources, that you can use to compete against the scale of Amazon. I think that’s the only solution out there.

David Mayer 11:53

This is the biggest difference between you two, I mean clearly OnApp, you’re building a huge federation of your customers, who are then able to–

Ditlev Bredhal 12:03

All customers. It shouldn’t be an OnApp thing. A closed marketplace is not a marketplace.

David Mayer 12:09

Just run us through the thinking behind that, and perhaps give us a flavor of where it’s going. I know you’ve got big plans that you can’t announce just yet.

Ditlev Bredhal 12:17

It goes back to the pool of resources. If we think about why Amazon is winning, I know there’s been plenty of Amazon bashing, but I’ll just continue on the same line here. If I asked all my clients why they chose to host with us, they’d probably say it’s because of uptime, or the support, or the price. If you look at Amazon, the uptime is questionable at best, the support is non-existent, and when you do the metrics, it’s actually really expensive. They live in a parallel universe, but that doesn’t matter. Somehow our industry, the hosting service providers, you guys; you totally screwed up, you slept in class and let this bookshop take over our business. It’s amazing.


Ditlev Bredhal 12:59

From that perspective, I think you have to look at why they are winning. From my perspective, it’s scale. We discussed that, it’s the case of UK2 losing the retail client, it’s geographical reach. They are in 11 data centers now, and they intend to open a new one every quarter. How can we, any of us, compete with that? The last part is products. I think if you take you guys, the 20 people here in this room, imagine we thought of a larger room, imagine we went to a conference of service providers maybe from 80 countries. OnApp, we have 200 deployments across 84 countries. Imagine if those guys worked together. From a scale perspective, Amazon wouldn’t have anything on us. From geographical reach, we would outperform them by what, 1,000? Obviously from a product base, all the products that are built on the OnApp platform, if you managed to federate across those, there would be one set of APIs, one UI in front of all this infrastructure, and we feel that’s the way to deal with them.

David Mayer 13:56

Tony, you’ve got quite a different take on things here.

Tony Lucas 13:58

Yeah, I think that sounds wonderful in the utopia that’s a happy hippy commune, basically–


Ditlev Bredhal 14:03

We’re doing it though, right?

Tony Lucas 14:04

No, I’m not saying it’s not technically possible, what I’m saying is there’s more downsides to it than there are upsides. Just to come back on your three points, it was scale, geographic reach and products. Amazon are only at scale because they had the cash behind it in the first place, which made it easy to finance, secondly because they’ve had the growth there to do it. They haven’t gone out and built at scale well in advance. If anything, Amazon could go out and build – you were saying one data center every quarter – they could probably open one data center every month if they wanted to. You look at Europe now, they’re still only in Dublin. They’re not in the mainland UK, they’re not even in mainland Europe. Clearly, geographic reach and scale is not quite as important as people make out. There’s all these people happily using it in Dublin–

Ditlev Bredhal 14:53

I lost a million dollar client to it, because of scale.

Tony Lucas 14:56

The scale point is valid, that was a really good point that came up yesterday, which is you have your key customer bases, you have the customers that you’re best suited to deal with, we’re talking about niches again. In a past life when I was in a host environment, I had one of the major motor racing organizations come to me and say the exact same thing, basically they’d like this much capacity for three days a month and this much capacity for 28 days a month, and I said Well, I’ll charge you for this then. The federation idea, there is merit in being able to federate workloads, but I don’t think it’s sustainable long term for two reasons. One, it’s all well and good having consistent API, consistent product set, consistent blah, blah, blah, how are you actually going to differentiate? How are you actually going to compete with all these other companies? Then finally, there is I’m sure spare capacity in service providers, but actually when they’re going to understand and monitor their capacity better as they grow, those margins, they’re going to be deliberately pushing to tighten those margins as much as possible. The law of averages says the bigger they grow the peak work levels become easier and easier to manage. Because you can redeploy the workload, which you can’t do with dedicated. So I think it might work in some use cases, but I think it’s overhyped compared to reality.

Ditlev Bredhal 16:11

We’ve been doing it for eight months, we’ve got more than $10,000 going through it every week, and it’s working.

Tony Lucas 16:17

$10,000 every week, how does that compare to Amazon? Even if it grows 100% a month, how’s that going to compare to Amazon in three years?

Ditlev Bredhal 16:27

It pretty much does by the way. The true point about being commoditized, saying infrastructure is commoditized is like saying all cars are commoditized, because–

Tony Lucas 16:37

It’s not yet, it’s not yet.

Ditlev Bredhal 16:40

How do we differentiate? The same way that all cars have four wheels but aren’t commoditized, by adding SSD drives, by adding fusion IOs, dealing with all the stuff on top of it. I really think our job is not actually for each of our platforms, it’s not to help our clients to de-commoditize themselves, because if we did they would be in effect commoditized if we provided them the same solution, right? They need to go into the niches, they need to do what they do best, not in the cloud, but they need to do their best according to what they have done prior to the cloud. They have all this goodwill they’ve built up in decades of doing business outside of the cloud, that’s what they need to focus on.

Tony Lucas 17:14

I have to ask you, Ditlev, where is the federation going? You’re going to be building– I know there’s a limited degree to which you can expose this, but for instance what you’ve done with CDN is you’ve made both your own CDN product and a white label product behind that which they can also use. You’ve already said that compute is coming up, where is this going and how good is it for the service providers in that federation?

Ditlev Bredhal 17:47

The end goal here is to provide an informed marketplace. What I was concerned about is not– when I started thinking about this federation was whether there would be enough demand here. I think that’s the real value of [inaudible]. Let’s say somebody asked me in UK2 back when I ran that, “How’s your hosting?” What I would say is it’s really good. What it boils down to is no-one knows until they actually start deploying it. Now in a federative setup, there would be an informed marketplace where the objective information supplied by the hosting company – all the uptime, the throughput, the IOs available, all the key KPIs infrastructure would be served to you as a buyer. Finally, there would be the subjective feedback from the other people using it. It’s the first time ever, I think, that you as an infrastructure buyer actually know what you’re buying before you pay for it.

Tony Lucas 18:36

That market is potentially there, I agree, but I’m personally of the opinion that it’s a market better sold by the likes of the RightScales, all these companies focusing on multi-cloud, because they’re the ones focusing on the business value in that area. The challenge with all of these is how does the service provider differentiate but still keep the customer relationship, and take advantage of it?

Ditlev Bredhal 19:01

The federation is really the–

David Mayer 19:04

Okay, we need to wrap this up now, but one minute from you.

Ditlev Bredhal 19:07

We think that the federation is the entry ticket, it’s not what you need to focus on. You need to focus on providing a service that can compete with Amazon, otherwise you pretty much have to give up or start reselling Amazon instead. We think that those three advantages, the scale, the geographical reach, and the products, you can’t build them yourself. I was fairly big business in my old setup, I couldn’t build it, I didn’t have the money to do it. If you if federate across, work together like the people in this room, we actually have a chance of saving this industry. If we don’t, we will look back at these years as the years when the long tail of our industry disappears, just like you’ve seen in pretty much any other industry that is scale based. Look around you, any other industry in the world pretty much has five players ruling it. The cases where it hasn’t happened is in logistics, that’s the federated approach. When you send something via UPS from the UK to the US, it goes through a thousand hands, that have nothing to do with UPS because it’s federated. That’s how to save the industry, and if we don’t, we’re screwed.

David Mayer 20:04

Okay, closer from you, quickly.

Tony Lucas 20:06

I think the benefits of the scale aspect there are overrated. You’ve got a hundred companies that have all got 30% spare scale, spare capacity and they’re working together, whatever the number is. They’re not going to achieve Amazon’s cost of scale, what they are going to do is reduce the downside of their spare capacity. That’s not a way to grow for me. It might be a way to survive, but it’s not actually a way to grow a business and actually be able to find out that niche and carve out and build the higher margins, it’s just a way to keep paddling along until you gradually exit somehow. Some of these guys may be sitting there thinking how do we grow, the answer might actually be don’t. Merge, partner up with other service providers instead, for some of them that is unfortunately the better idea.

David Mayer 20:52

Okay, we’re going to have to leave it there, but it’s been absolutely awesome getting you guys on stage, I’ve been looking forward to this for a while, so excellent, thanks.


firstpage of 2
  1. I think it’s important to sometimes look at what has happened in other industries. Walmart took the USA by storm in the 80′s and 90′s. It served everyone — and with their plan, there was always one nearby. All the mom & pop shops quickly went out of business, even if each of those places had their own strategic advantages, whether it be customer service, product … etc.

    In cloud hosting, it’s without question that Amazon is Walmart, and every other cloud host are the mom & pop stores. Ditlev is absolutely right; unless cloud hosting companies work together, the same thing is going to happen.

    1. Terry,

      Think you might need a disclaimer on there on who you work for… :)

      (Ditlev/Onapp for anyone wondering).


      Tony (Disclaimer, yes I work for Flexiant)

  2. Nice speach Ditlev and i totally agree.. Look at UPS!!! haha.. you have used that before and i totally agree!!!


Comments have been disabled for this post