WANdisco Changes the Game in Distributed Computing

WANdisco, a 25-person startup headquartered at 4847 Hopyard Road, has released a series of products over the last year that help development teams deliver on the productivity promise of offshore development. The savings suggested by being able to develop software globally at multiple sites connected by a wide area network (WAN) are impacted by the inherent difficulties in coordinating the output of teams of programmers located at multiple sites around the world.

For example, failure recovery and backup are time consuming. Synchronizing the sites can be complicated and slow, involving a central transaction coordinator that updates a master repository - the "active" repository - which must then be replicated at the other sites, and once developers start working again, the sites no longer match.

An ideal solution would be to have each development team, no matter where it was located, work from a data repository that was identical to and synchronized in real time with every other data repository, something referred to as active/active replication, and once thought, according to theoretical mathematics, to be impossible. Rahul Bhargava and Yeturu Aahlad thought otherwise and founded WANdisco in 2001 to prove it. Considered one of the most difficult problems in computer science, Aahlad likened it to figuring out how five people at a dinner table can eat simultaneously with four forks.

It took three years, but Bhargava and Aahlad solved the problem with some complex synchronization technology, a new network protocol, and a sophisticated application that lets development teams around the world work as one unit in real time. Each site has an exact copy of all of the data and is always in complete synchronization with the others.

WANdisco's first suite of products provide active/active replication for CVS and Subversion, two popular products used to manage source code.

"We decided that the first market for us would be source code management systems," says David Richards, WANdisco CEO. "Some of our customers have up to fifteen different offshore development centers. Their developers can now work together at LAN speed, as if they are communicating with a computer that's right on their desktop. It used to take developers a day to check files in and out. Now they can do it in about 2 minutes."

WANdisco's multisite real-time synchronization is so effective that disaster recovery happens automatically, requiring no action from a system administrator. When a network connection or server fails, developers just continue working. When a line went down between AT&T's U.S. headquarters and a development center in India, developers barely noticed, and when the line came back up, the software synchronized itself automatically. No work was lost; down time was zero.

"Normally with startups, the first set of customers are small companies willing to take a risk," says Richards. "But our customers are very large companies, billion dollar public companies, using our software to manage their single biggest asset, their source code. Yet every one of them, when we first talked about this sort of performance, said it wasn't possible. Until they tried it. This is probably the easiest piece of software I've ever had to sell in my life."

Photo: CTO Rahul Bhargava, Yeturu Aahlad, PhD, and CEO David Richards have solved the synchronization problem that once faced distributed computing.

Also in this issue...

Share this page!