EBay's computing guru gives behind-the-scenes peek

Today on eBay, you just might find that absolute perfect mantelpiece you’ve been looking for, at a great price. It’s there, nestled among some 100 million other items, placed for sale by one of the online auctioneer’s 233 million registered users. Now think about the back-end infrastructure that enables you to find, and then buy, that object of your delight, and you do have to wonder how it ever happens. Contemplating the database environment alone — 600 production database instances spread across hundreds of medium-sized servers — is enough to give even the most stalwart IT executive a case of the shakes.

But Paul Strong, distinguished research scientist at eBay, doesn’t faze easily.

“You just can’t get all the details for 100 million items on a single machine,” he said in a recent interview, during which he described eBay’s IT infrastructure, discussed next-generation trends, and shared how any enterprise, large or small, could benefit from the lessons the online auctioneer has learned along the way. Here are some of his thoughts, which he’ll also share during an opening keynote address at IDG World Expo’s joint Next Generation Data Center and LinuxWorld conferences in San Francisco Aug. 6-9.

The next-generation data center today …

When we look at the data center, we don’t see silos and silos of applications on islands and silos of infrastructure because those have proven to be expensive and not particularly efficient, and they tend to be very static. We need to move toward [something] more dynamic, and that means really viewing applications and business services as being network-distributed. And the platform on which they run is the data center. The data center is a system and should be treated as such. The application components are distributed across the entire system. How your application behaves depends on where your load-balancers direct traffic, the number of application instances behind them, how you connect to your databases. Your applications and services don’t run on a single server. They run on a collection of resources that range from servers to firewalls, load-balancers and such like.

Where the next-generation data center is headed …

One of the real trends in the next-gen data center is that it’s all about interconnectedness. It’s about the fact that all value is delivered by connecting sets of things together and agility is achieved by reconnecting the same sets. So it’s all about relationships and how you manage them. It’s the relationships that deliver value and how you cable together your infrastructure, how you make your applications and services communicate, and the patterns you use to drive the value it delivers for the business.

Managing the next-generation infrastructure …

We’re using some technologies, for example semantic Web technologies, to allow us to have an ontology that describes our infrastructure and allows us to ask questions of it. We want to be in a position where we can ask our management framework, ‘If a user presses this button, show me the things in the path.’ And if they have a problem with it, ‘Show me everything in the path that could be broken.’ Or if, say, a load-balancer in our infrastructure breaks, ‘Show me which business process is impacted so I can understand the financial impact on our business.‘ Things like that.