Limit your financial exposure
CRN: In building their own infrastructure for customers, resellers are taking on a lot of risk and the market is moving very quickly. It makes it difficult to know which platforms to choose. How did you make your decision and what did you take into account before you made it?
Somerville: You can only make each decision based on a business case. So each time you make an infrastructure decision you sit down and you make the evaluation based on what opportunity is sitting there as well.
I would suggest most of the integrators who are moving into that space would be making more pointed solutions, as opposed to going, I'm going to build a big cloud infrastructure and go to the world and say here it is, come and use it. We're building infrastructure [where] we know [it] is specific to a set of customers or a customer base that fits their need.
CRN: And they've signed a contract for that?
Somerville: Well, no, sometimes we build it before that. But also it's about building a platform that's scaled...
Illot: It doesn't matter when you build it, it's got to be customer-led. You need to be providing something that your customers want.
Bertolini: Do you include power and cooling in your solutions?
Somerville: It depends - if we were to go and take rack space in somewhere like Global Switch, we don't need to worry about it if we're just buying by the rack. But if we go and build infrastructure in our place, which we're doing, then yes we have to take that into account.
And again, we've just been through the process of doing the numbers. I can go and get 10 racks here or I can build and put 10 racks in here. Do the numbers. It's actually an interesting thing to go and do...
CRN: You say the numbers are interesting - what's interesting about them?
Somerville: Well a rack in a data centre is one thing, a rack with 10 kilowatts of power in a data centre is another. That's the reality.
If I use an HP example, a C7000 [HP Bladesystem] and an EVA [HP disk array] can use 10 kilowatts in a rack. So if I'm provisioning four racks to run a managed service platform then I have to assume I'm going to fill those racks, so I need 10 kilowatts a rack.
Then when you start looking at the economy of 10 kilowatts per rack in a data centre today and then do the numbers of building the data centre and you look at the per month cost, it depends on who is doing the pricing and where you are looking at your floor space and all those things to go with it, it's not necessarily forgone that you go and stick it in a data centre.
Bertolini: Can the banks help?
Somerville: They can help by dropping their interest rate! (laughter)
CRN: I'll let them know.
Somerville: Yeah, can you do that?
CRN: There's an interesting point there where you say it's customer-led but not necessarily covered by contract, so you are still taking some risk. What if a cheaper technology came along next year?
Illot: That's the nature of our industry; it doesn't matter what you buy, it's always going to be cheaper in 12 months' time, and faster and better and bigger and all those things. You can't wait forever...
Somerville: Everything we do as an integrator in this space is a risk. We sign with a vendor and decide we're going to become a partner and we invest 50 grand in training certification and all that. That for us is a risk and an investment. If we don't go and pull the revenue to pay that risk, that's what we do and that's what we've always done and this is just a different flavour of the risk that we take...
Illot: If you don't take those risks and make bets every now and then, hopefully educated bets, but if you're not doing that every now and then your business doesn't grow, it's as simple as that.
CRN: I read a blog recently about a consultancy called Cloud Scaling which builds cloud infrastructure for enterprise and ISPs like NTT in Japan. The blog talked about the commoditised cloud versus the enterprise cloud and said that their approach using commoditised IT is as effective as enterprise IT and yet cost a tenth of what you'd get by doing it with Vblocks, for example.
McCullum: Bad example. (laughter)
CRN: Yeah, I wasn't going to use it.
McCullum: But is he comparing capital costs, like what it costs to acquire, because he's sunk a lot in R&D in investment in building that out himself?
CRN: I think he just tells them what to spend...
Prosper: In that sense it is commoditised hardware, it's a low-cost, zero-cost hypervisor on the Xen platform, typically with a Linux platform running on top of that. So you start looking at the cost of that and it's actually quite inexpensive.
Lamas: The problem is that was a very broad statement. Where are the numbers and then we can talk because maybe we are talking about different things.
CRN: The theory seemed to be that if you have many cheap servers it doesn't matter if they fail because you make it up by the numbers, versus buying a best-of-breed approach.
Lenz: The other question is, have they done the old "build it and they will come" model, or is there a business case around what's actually being put out there today? I think you'll find a lot of people building cloud services today have really got a fairly strong business model, but you've got a lot of people who invest a lot in a truckload of iron and views that people are going to come to this type of thing, and I think there's a balance...
Somerville: It's really easy to fire a virtual machine up. You want a virtual machine with 100GB of storage, no worries, there it is tomorrow. Replicate the data under it, sign a contract to say if I lose it I'll bleed, and suddenly cost is associated with it.
There'll be tonnes of those offerings out there but if an organisation says my business runs on my data, prove to me that I'm not going to lose it, there's a cost [to] that.
It's not that they can't do it but that would build the cost of that service. You can always go and get something cheap - you can go and buy a cheap motorcar, no worries - but is that delivering your requirements and expectations?
Lamas: The cost curves cross. Maybe it's cheaper, I would like to see the numbers but I'm sure different inflections occur. They are going to meet at some point, maybe in one year or two years' time, one is going to be cheaper than the other.
CRN: I was interested to see that Tata in India has just started its own competitor to Amazon Web Services. I thought if they're charging in rupees it's going to be interesting, but actually the numbers are almost exactly the same as Amazon.
McCullum: That goes to my point earlier on - as vendors and partners [we need to be] working with customers to define what the requirement or success looks like.
If success for a customer or partner is purely price and McVMs, swipe your credit card and get it and trash it when you don't need it anymore - if that's success then go and do it.
If success is high levels of availability and predictability and a range of other different characteristics then that would drive different decisions and different purchasing potentially.
Hume: I would suggest that an Australian channel is quite mature, so price is less and less a consideration for the enterprise. So it's about agility, speed, flexibility and aligning that to the business...
Next: The advantage of being local
-1.jpg&w=100&c=1&s=0)
.png&w=100&c=1&s=0)
