Imagine creating and managing a datacentre that handles 241 million registered users and sells a car a minute, an auto part every second, and a piece of diamond jewellery every two minutes.
Imagine a datacentre that changes its code every two weeks and executes 20 billion SQL transactions daily.
This was the subject matter of the keynote speech, Examples of Next Generation Data Centre Technologies at Work, delivered at the LinuxWorld/Next Generations Data Centre conference in San Francisco by Paul Strong, distinguished research scientist at eBay.
Strong led off his address by saying he preferred to talk about the next-generation datacentre in terms of the value it should deliver to the business rather than in terms of technology.
"Datacentres are about running business processes driven by SLAs," said Strong. And, he continued, datacentres can become "value centres" rather than cost centres when IT and business communicate.
To that end, companies must decide what is core to their business and what can be commoditised as a utility served up by a software as a service (SaaS) provider.
Sighting eBay's huge volumes, Strong described how its solutions will help future datacentre users scale their growing infrastructure.
In the case of eBay, it faced its biggest challenge back in 1999 when IT executives realised the database couldn't scale any further.
"The database was approaching the limits of physical growth," said Strong.
The resolution came about when engineers were able to virtualise the database by separating the interface from the implementation. The interface remained constant, but the implementation underneath it could change by decoupling the database.
Another change to the infrastructure came about when eBay was able to program the database in such a manner that it behaved differently depending on what command hit the application.
"It became like a giant field-programmable gate array," Strong said.
The next major challenge that eBay faced was how to track and manage the hundreds of thousands of interdependent relationships within the application services. For example, Service A might be composed of subservices B and C, while C might be composed of two additional elements.
"We are beginning to categorise these relationships," said Strong, and over time, eBay hopes to build in business rules that can automate and manage what can no longer be managed by any single group of IT personnel because of its scale.
Strong ended his talk by telling the audience to think of the datacentre as a machine that runs business processes.
Following the keynote, Strong was asked if he thought the fact that Linux and open source were never mentioned by either him or the preceding keynoter, Werner Vogel, CTO of Amazon, had anything to do with the diminishing importance of which operating system is used by applications.
Strong did not see that relationship and said Linux continues to be a major component of eBay's infrastructure, but he did admit there are other issues in a datacentre that are taking precedent over which OS to use.