Jun 20, 2008

One Chapter Ends

Three and a half years ago, I had joined a new start-up called Cassatt. At the time, I was working in the Java Web Services group at Sun when I got the call from Cassatt's recruiter. I wasn't closely following Cassatt back then, but it was on my radar. In late 2003, I remember reading about Bill Coleman's departure from BEA to form a new enterprise software start-up. And then in April 2004, Rich Green left Sun to join Cassatt. When I joined Cassatt, I didn't fully understand what Cassatt was trying to do, but I knew it was developing some cool, transformational technology. Bill has this saying "Go big or stay home." Well, I don't ever remember staying home during my time at Cassatt.

When I joined in October 2004, we had just started developing the Collage 3.0 product release, with the codename "Armstrong." The longer codename was "Armstrong, the moon guy, not the bike guy." That was one of the years Lance Armstrong won the Tour de France, but the release was named after Neil Armstrong. After shipping Collage 3.0 in February 2005, we took that one small step and firmly planted our flag in the sea of utility computing. (Remember, it's a moon metaphor.)

Over the next 15 months, we released three more versions of Collage. In January 2006, Collage 3.2 "Apollo" introduced support for Solaris. In April 2006, we shipped Collage 3.3 "Spyglass" which introduced support for managing virtual machines. Along the way, we ended our era of space-flight codenames and started a new era of famous golf courses. Okay, you can aim for the stars, but as far as I know only Alan Shepard has teed it up on the moon.

After the Collage 3.4 "Starbuck" release in February 2007, we transitioned to a new development process. Large releases were replaced by small projects that could integrate to Head of Tree asynchronously. There were two invariants: Head of Tree was always shippable, and projects integrated when they were complete and bug-free. At anytime, there were about a dozen projects "in flight." (Okay, we didn't get away from all the aerospace metaphors. We have too many pilots in Cassatt for that to ever happen.)

This was a pretty significant change in how PD operated, but the results were great for the company and the customers. In 2007, there were over 80 integrations to Head of Tree, and the time to deliver new features to customers was dramatically reduced. There was a renaissance of product functionality. We added support along many dimensions: new hardware from Sun, 64-bit support, Xen VM's, new VMware versions, additional networking support, performance enhancements, stability and robustness.

And then in November 2007, we launched the new Cassatt Active Response 5.0 product line. While supporting our utility computing mission, we added support for power management of engineering labs and dev/test environments. Since then, we've continued to add features incrementally. Just recently, Cassatt Active Response 5.1 introduced demand-based policy that should prove very helpful in power-management and utility-computing environments.

So as I leave Cassatt, I leave with mixed feelings. I've enjoyed my time at Cassatt and look back at fondly at our accomplishments. I will miss all the talented folks I had the chance to work with. But I know "it's a small valley" and our paths will cross again.

Jun 11, 2008

Desktop Virtualization

Check out my new column at Virtual Strategy Magazine. I will contribute to their new "Insights" section, which features contributions from several other authors. This month, I talk about why desktop virtualization is a bad idea (and why it's not new either). (Read all about it)

Let me know you what you think.