One of the first questions I think everyone should ask any vendor who brings forth a new offering is whether that vendor uses it in-house. I've worked for a number of technology companies. Some did and some did not use their own technology and I've generally found that those that did focused on customer needs more tightly. Those that didn't covered up problems more effectively. You can't cover up problems very effectively if the stuff is breaking in your own shop. We call this use of your own stuff "eating your own dogfood."
A few years ago, HP started to use its own technology internally -- aggressively. This change has had a profound effect on the kind and quality of products it brings to market. One of the most telling is the Converged Infrastructure effort, which is targeted at large enterprises and midmarket companies that have aspirations to be large enterprises, plan to grow by merger or be acquired, and/or have a need for enterprise-class data centers. Let's explore HP's Converged Infrastructure.
The Data Center Problem
As most of you who run, build, or own data centers likely know, the biggest problem you are generally dealing with is a lack of common standards. This ranges from hardware to software and cuts across technology types like servers and routers. The data center of today is an ugly mess of stuff that was never designed to go in the same room together, let alone interoperate.
Now you have to cool this entire mess, and the end result is that the cooling system has to be massively over-engineered to power though migrating hotspots and around bundles of wiring in what often is an airflow nightmare. The data center turns into the wind tunnel from hell. And then it ends up like a shanty town that grew up as capacity was needed without any planning commission or architectural guidelines to assure safety or consistency.
Worse Under Merger
Now when you plow two companies together, you generally centralize on one of the two firms' platforms, which means you suddenly have to change a lot of stuff at once and grow the data center. This is often like trying to add floors to that shantytown. The foundation sucks, so the only way to really do it is to yank it all out and start over. But that is excessively expensive and time consuming, so the systems are integrated for long periods of time and many of the benefits of the merger aren't realized until much later than expected.
This became painfully obvious to HP, both when it acquired Compaq and again when it acquired EDS.
There had to be a better way.
Building for the Future
The idea was to create an infrastructure that could be easily changed and grown. For that, HP needed something that was modular, open, could be planned, and yet still was as robust (if not more so) than what was being replaced. The benefits were believed to be faster time to revenue, lower costs associated with acquiring and implementing new hardware, much higher flexibility in being able to respond to business changes like mergers, and one overarching benefit of lower risk, which comes directly out of much of the rest. The effort was named HP Converged Infrastructure, and I was recently briefed on it. It spans servers and networking, it includes energy management, it includes anticipatory wiring (so you wire once), and it is designed to be effectively managed centrally.
HP and Building to the Benefits of Modular Hardware
One of the most interesting aspects of this modular HP approach to hardware is the use of massive imbedded sensors to optimize cooling. This allows the same rack structures to house a mix and variety of server types and networking hardware and allows all of it to be changed out by changing a shelf and some modules rather than replacing the entire cabinet or rewiring an entire floor. This includes technology ranging from Integrity to Superdome servers and a variety of server types, including blades.
The goal HP achieved is that it needs only to rack once, wire once, and power once. Changes are close to plug-simple as a result. It also includes integrated networking, and this is likely HP's strongest response to Cisco entering the server market. This includes everything from routing to threat protection like firewalls and intrusion prevention hardware like the Tippingpoint solution. This is also the first visible integration of the technology it acquired from 3Com because it will integrate 3Com's enterprise-class data center switch.
Finally, it includes storage modules like Scale-out SANs and Scalable File Systems able to scale up to a massive 16PB for a single namespace.
I'm not aware of anyone doing anywhere near this level of standardization. While it would be great if this were multi-vendor, just doing it inside one company is farther than anyone else has gone.
Wrapping Up: The Power of Eating Your Own Dogfood
As impressive as this effort is, the core point is that it likely wouldn't have happened at all had HP not started aggressively using its own technology. This is because divisions often don't want to cooperate to the level needed to drive cross-company standards. HP is embracing open standards with regards to software and solutions, but at least inside the company it is standardizing on hardware. That likely makes this solution stand out favorably against those from multiple companies or from one company that lacks this level of hardware standardization.
In the end, it goes back to the question I recommend others ask of vendors -- whether they aggressively use what they are selling. If they do, their products will generally be better. If they don't, maybe you shouldn't either.