With the summer recess about to start, we caught up with HP's UK server and storage manager Iain Stephen, to get his view of how well HP is doing, especially when it's up against aggressive competitors such as Sun and Dell.
Q: Do you see much competition from Dell?
A: They probably are, but our channel at that end of the market is very indirect. They have a good portfolio for backup services, and more and more they're using small value-added resellers and paying them flat fees to fulfil their contracts.
The mix of sales is shifting more and more towards racks - it's now 85 per cent rack-based. In small companies, racks aren't as unusual as they were, it's a form factor that drives better use of the product. You still see towers on shelves but a couple of years ago we ran a campaign to covert towers to rack-mounted for free. It looks neater - they've really stretched away.
Q: What about standardisation?
A: Each vendor thinks they can add value by adding their own features so no-one will standardise right now. Most customers accept that and having made their choice, they quite comfortable with it. Our blade portfolio is one-way, two-way and four-way. Most vendors have given up on the one-way and four-ways and concentrate on two-ways. There's lots of applications for one-processor blades.
Q: What about management?
A: Management tools - there's a general nervousness that even when it comes to rack-mount servers, you can go and plug something into them. With blades we have diagnostic tools which you plug into the back of the shell - it's a bit clunky. It's the hands-off nature of blades that make people nervous. They want a USB port to plug in a keyboard into the front panel - now the whole idea is to do away with sockets to make it smaller. I think that attitude is gone.
Q: How do you define a blade server - that is, what pieces go on the blade and what in the chassis?
A: The optimum is that only chips, chipset and RAM go on the blade. Ultimately, no storage, and only Fibre Channel out the back. Storage is separate, and there's limited memory requirement and remote intelligence. The rest is supplied by the chassis. You can boot these things from the SAN now so most people understand what they are and how they can use them. They can make the link between hardware and application.
Q: What about the cost of licences?
A: Any software is expensive compared to hardware but one of the big question marks is how to save money. Linux people ask this but the Enterprise Edition costs money too and even then they only save on maintenance. You still need as many people to run it and they can be more expensive than the people they've already got. Most have two worlds: an NT world and a Linux world. They could go down one route or another but haven't. Maybe they haven't felt like sacking half their IT dept.
Q: What's happening to HP/UX?
A: It's not on the way out - it's one piece of our added value and as long as people want to buy it from us we'll keep enhancing it. There's still strong demand as does OpenVMS. But when it comes to large SAP bids it's competition between AIX and HP/UX.
Q: Why do you need big Unix boxes?
A: SAP at the back end is the kind of application that lends itself to big box architectures. We’ve been doing proof of concept on AMD boxes and they're performing better than Intel boxes.
These days people don't need the huge headroom they used to always be chasing - the places big boxes still play are SAP, Oracle, and anyone who's coming for a big box environment and who wants to consolidate. If they go for a 128-way SuperDome they can get rid of 32 machines. There's a lot of that. In a few years time they'll go for a lower cost topology but right now it's about consolidation.
Q: Does clustering cut it in this environment?
A: It depends on the application - for Web serving, Microsoft Exchange for example, it works. Right at the top we've got fault tolerant boxes and they're not just for financial services companies.
Newspaper publishing is a big customer for non-stop computing, they're paranoid about losing their copy. But until two years ago the main buyers of non-stop computing were telcos, big banks, investment banks especially the Link exchange. But now because of software from a company called IDX, fault tolerance become popular in government. There are 34 non-stop customers in the UK - every now and again a new app comes along that allows you to go and find a few new customers.
Q: How do you see IBM's efforts?
A: They're doing rather well but the one that isn't is Sun. We tend not to see them because their high end market share is mostly in the NHS. Most of those projects have gone live now so it'll be interesting to see how well they do in calendar Q2 - just to see if it just stops completely now. But mostly, it's us against IBM.
Q: Can you write Sun off?
A: At the high end, manufacturing and retailing are very risk-averse. That means Sun doesn't make it to the short list any more because of the uncertainty surrounding them. The bit people love is their environment - they could take better advantage of it.
Q: Should Sun take Solaris open source?
A: If it were me, I'd move their OS onto another platform and charge $999 for it, which is what everyone else is doing. They're in the same position that Digital were. Everything in their environment is based around RISC technology - it's high value, high margin yet they're trying to compete with IA32-based systems at a thousand pounds a pop. It doesn't work - eventually they're going to run out of cash. They should go very software-driven.
Q: Is the long term future smaller boxes? For example, Google manages on thousands and thousands of cheap Linux boxes.
A: It depends on customers having the capability for it - for most it's too far out for them. This is where the Linux and blade thing comes in, for example for financial customers.
Q: But it's a very scalable model.
A: Yes, but go back a few years when everyone was client server - keep adding things and it'll work but I don't think most people are comfortable with it because it's fear of the unknown, it’s too high-risk, and a lack of everybody else doing it. The answer you get back if you suggest it is that 'we're not a research institution'.
Q: You launched an Opteron box recently. Are customers asking specifically for AMD chips? Why wouldn't they want a 64-bit Xeon?
A: Investment banks and research companies, telcos, are buying Opteron boxes to see what they can do with them, and it was the only 64-bit chip around. We needed something to give us a short term stopgap until the Intel product arrived.
Q: You mean, it helps you leapfrog Dell?
A: It gives us something no-one else has got, everyone's got a two-way, we've got a four-way which people are more interested in. The other advantage is that it uses a different memory subsystem which is faster. So there are other advantages to the Opteron design which is not a traditional Intel chipset product. It uses DDR2 and other, faster I/O systems. But for us it's just 500-600 units a month out of 9,000.
Q: Are there compatibility issues with AMD?
A: No, it's to do with the amount of sway the Intel name has in boardrooms.
Q: How do you see the market developing, generally?
A: Analysts talk about double-digit growth, Unix boxes are pretty flat. Storage -- though there's a huge demand for capacity, revenues are staying flat as prices fall so for us vendors it's pretty tough.
In terms of the competitive landscape, everyone's looking at Sun's market share, especially in file and print. Their prices are pretty equivalent to Intel machines but we offer expertise and software.
Client consolidation is another trend we're seeing - thin clients are coming back, especially in places like call centres. Price points of servers are low with the emergence of VMware - people are experimenting to see how many server images they can run on one box.
The issue there is OS licensing and whether Intel and Microsoft will agree on whether the new multi-core processors are single or multiple processors from the point of view of licensing. That's a huge issue.