Dell cooks up new HPC strategy
Not just a box-slinger anymore
HPC blog As I trudged toward a swanky hotel for a meeting with Dell, the Seattle sky was spitting cold rain like an old man realising the soup in his mouth is way too hot. (Adding more drama to these intros, nice, right?)
I expected two things that day: “It’s Seattle in November; it’s going to rain,” and “It’s Dell at Supercomputing; they’re going to talk about hardware.” Only one of those assumptions was correct.
Instead of talking hardware, the meeting was all about Dell’s HPC strategy and how it is going to engage the market. It wasn’t a typical Dell-like meeting, where they’d reel off server names and configs and I’d nod appreciatively, “Hmm... so you’re going to put newer/faster processors in that one? Way to go, nice job...”
Dell seems to have put a lot of thought into its HPC strategy; it is more focused and nuanced than anything I’ve seen from the firm before. The overarching concept is that Dell wants to be the HPC vendor of choice – not for the Top500 but for the “other 5,000".
This doesn’t mean it won’t go after Top500 opportunities, but that Dell is going to carefully pick its spots to make sure that the deals it pursues aren’t unprofitable death marches into the unknown bleeding edge of supercomputing. When the deals get huge, Dell wants the projects to be well-bounded with workloads and customers it knows and understands – like hometown TACC.
That said, Dell's main thrust centres on selling smaller HPC systems – the systems that the “other 5,000” buy. Dell believes that its economies of scale, efficiency, and maniacal focus on costs gives it the advantage in this market segment.
I agreed that this is the right place for Dell to concentrate, but this strategy raises a number of potential sticking points. Some of the characteristics of Dell’s basic business model don’t line up all that well with HPC – or even mission-critical enterprise system selling.
For example, Dell has a propensity to change hardware configurations with little or no notice. While swapping out a motherboard NIC to save a few cents won’t make a difference on a typical PC, it can make a big difference on enterprise and HPC servers. Sometimes these supposedly minor hardware changes raise hell with complex software stacks, causing problems that can be hard to troubleshoot.
Dell responded to this concern by saying that it has a new policy on the server side: full hardware availability for at least five years. Customers will be able to get the exact same configurations with no hidden changes, meaning that their carefully crafted processes and apps will run without question when they buy new copies of existing systems.
Perhaps the biggest hurdle for Dell is fixing its reputation for being just a box-slinger. Even mid-sized HPC customers look to their system vendors for help in selecting, sizing and configuring systems. This means the vendor needs to know a fair amount about HPC apps, how to tune them, and how to make them run best on their gear. It also has to have skilled personnel who can handle all of the above at the customer’s site – this isn’t the kind of stuff that you can cover in a few chat sessions with an online salesperson.
This hasn’t been a historic strong point for Dell. It’s not viewed as having a lot of HPC savvy compared to IBM, Cray, SGI, HP – or even smaller players like Appro. According to Dell, this is something that it has been working on steadily for the past several years, adding technical expertise both in the field and at HQ.
One of the ways it is combating this perception and helping customers feel warmer about Dell as an HPC vendor is with its HPC cookbooks. Dell is putting together recipes for popular HPC apps in small, medium and large configurations. Each of these is fully tested and guaranteed to hit the promised level of performance.
A recipe includes the most efficient hardware configuration, the software stack, and instructions for setup and tuning. Each recipe configuration can be altered to fit customer requirements or to fit in better with the existing infrastructure. With each HPC proposal, Dell includes the recipe plus pricing on the option to have Dell handle the tuning and installation chores.
It sounds as if Dell is on the right track with the overall strategy and approach. If it builds up a broad enough list of recipes and follow through on what it's promising, it could do some significant business in the “other 5,000” market.
But this isn’t going to happen overnight. Dell is going to have to provide a sustained investment, and commit enough time and money to get the job done. If it is looking for big results in the short term, it is going to be disappointed. It will take time for Dell to beef up its offerings and to change customer perceptions. ®