Last year was a good one for HP's server division, so when HP's VP of enterprise service and storage Paul Miller hit town, we took the opportunity to gain insight on how the company plans to build on that momentum in 2008.
Q: How do you see HP differentiating itself this year? A: I think what we are seeing is a change in the marketplace and HP is the best position to take advantage of it obviously. It’s no longer about just servers or storage; it’s about infrastructure and blades is probably the best positioning of that.
There's convergence happening in the marketplace, convergence of servers and storage, convergence in mission-critical x86 industry-standard devices, and the convergence of the interconnection fabrics. Those three things are starting to change the known culture of the environment.
When I say that it’s changing from a server storage to a infrastructure sale, it’s becoming a more complicated environment. We are seeing less reverse auctions, we are seeing more consultative selling because - one of the customers yesterday, his main problem is his datacentre is out of power and cooling.
You’re starting at the datacentre walls of the infrastructure and breaking it down and instead of you leading with the specs - all slots and watts - it’s the last thing we talk about. It starts with the power and cooling, it starts with interconnect then it goes down to the management infrastructure. We’re spending all our time talking about the infrastructure, not really spending time talking about specific in how many processors or how many slots are in a box anymore.
Q: So datacentre managers are saying they have a power envelope I have to work within, but I need to do twice as much? A: Exactly, that’s one of them, or "here’s my datacentre it’s completely out of power and cooling, how can you help me put more into it so I don’t have to build out their datacentre which is a tens of millions of dollars proposition."
It starts off with "I’m wasting time, my team no longer has time to go off and do simple maintenance and simple tasks. How can you simplify the tasks because I’m just in execution mode, I don’t have time to go off and invent and do new things, I’ve got to put out new applications." They want to spend their time on bringing new capabilities to their customers, not maintaining existing capabilities.
Q: But if systems are more reliable now then they were, what are they spending that extra time doing? A: It’s about software upgrades, it’s about changes. This is why I think blades have been so successful because we’ve simplified how you change servers and storage.
If you look at it, what’s interesting is virtualisation. They spent a lot of time last year bringing out virtualisation. It used to be that the average server utilisation was around 20 percent. Virtualisation raised it maybe to 40 percent is what we’re seeing, so it hasn’t done everything it needs to do. They’re spending time in getting more utilisation out of my systems, so you’ll see us be very aggressive in working with customers in building out products that help them increase that utilisation.
Q: Just to go down that road for a second, how far do they need to go because then obviously you need some overhead? I mean you don’t want to run everything at 80 percent do you? A: What you want is an environment where you are running everything as close to 100 percent as possible quite frankly, and 100 percent means with your overhead. The problem is how do you plan for that overhead, how do you make sure you have the right amount of overhead? And an ideal scenario says you have just enough capacity and you’re shifting it around so that you’re able to achieve that, that’s the way you are going to get more and more into your datacentre.
Now somebody will say "well that’s totally unrealistic." It might be totally unrealistic at 100 percent but that’s not a bad goal to have. If you look at most applications, they are periodic: they run hot and they run cold. So you want to be able to shift those resources to where they're needed.
The pressures I see in the datacentre are going to continue to drive it from 40 to 50 to 60 to 70 percent. If we ever get to 100 that will great. I don’t think we will, but I think it’s going to head in that direction.
Just talk to any datacentre here in London: they’re plumb out of utilisation, they’re plumb out of space, plumb out of power, yet they’re running their servers with storage at 40 percent utilisation. There’s a lot of opportunity to unleash opportunity capacity within this and I think we have the technologies.
Q: So what do you say to them when they say, "I need more utilisation," what’s your answer? A: We usually start with the datacentre itself, we talk to them about Dynamic Smart Cooling, today we’re the only ones delivering that.
Q: But that’s just one part of the service. A: We then ask what their enterprise management strategy is, we usually start talking about our blade architecture. It requires less air conditioning because of the jet fans that you talked about.
We then start talking about the management architecture that we have across the board. Blades need the simplest and easiest management so we spend a lot of time talking about how they automate and simplify server setup, storage setup, building out simply, so our tools around SIM, around Insight Control are really designed to automate and simplify maintenance and setup.
I’ve got to take time out of automating the distribution of software patches, those are big things. With the virtual connect, the ability to dynamically to continue to allocate more storage to a device. The ability to reallocate between storage and servers, driving those capabilities are very, very huge and big to customers, that is why blades continues to be for us double digit growth.
When we built our blade chassis we didn’t focus on the granularity. The second order was the granularity of the form factor, the first order was self-power and cooling, making it simple and easy to manage and then driving this new level of virtualisation.
Our virtualisation strategy is above the hypervisor. We support VMware, we’re the largest reseller of VMware, more is done on our boxes than anybody else's but we see a broader vision of virtualisation. It’s about the hypervisor layering on top of that, the virtualisation of the interconnect and then virtualisation management on top of that.
You'll see us being very bold and aggressive this year taking and automating that all the way up through our Opsware acquisition. So end-to-end virtualisation is what we see as our strategy, and we think that where customers are going to go.