Around 2001, my good friend Tim Cramer and I were working together on the Java Performance team at Sun Microsystem. At the time, Sun was launching a new computer server codenamed Starcat. The Starcat, later officially dubbed the SunFire E15000, was a massive beast. With a price tag over $1,000,000, it was the size of a refrigerator and it included an unthinkable 72 CPUs on a single, coherent backplane.
Our job was to make the Java Virtual Machine (JVM) work effectively on this system. We spent months tuning and tweaking before the launch of this machine, and I look back on that time as one of the most fun times in my career. This was my first exposure to running software at massive scale.
Several years later, around 2007, I got involved with another project called Ranger. Ranger, at the time, was being built by the University of Texas to be the world’s faster supercomputer. It consisted of around 4,000 physical computer systems, networked together over a super-fast Infiniband network.
While I initially thought of Starcat as massive, being the size of a large refrigerator, Ranger was closer to the size of a football field! It was designed to tackle the worlds largest scientific computing problems like weather prediction and nuclear fusion simulation.
Managing this beast as a coherent set of resources was massive undertaking. We used software like Ops Center and Grid Engine. We had to push both of these products to their limits and it was a huge learning experience. It took us months of experimentation to really make it all work together.
More recently, I joined Citrix and began working on Citrix CloudPlatform (powered by Apache CloudStack). CloudStack is software that helps customers provide Infrastructure as a Service (IaaS). IaaS allows users to generate new computers, complete with storage and networking using various virtualization technologies.
In essence, it allows users with relatively basic system administration skills to create and manage massive virtual data centers. Since I joined Citrix, I’ve started to meet customers who are running systems that are 10 times larger than Ranger. The scale keeps increasing.
This week, I’m at the CloudStack Collab Conference in Amsterdam, where we’re bringing together the community that builds this amazing software. This week we’ll be talking about where we take this software and what comes next.
While there are many sessions for people interested in different topics, for me, one thing that remains interesting is this idea of massive scale. What comes next?
One goal I was discussing with one of the engineers recently was what it would take to create a virtual data center with upwards of one million virtual computer servers in just a single day. It’s a big, hairy, audacious goal, but there’s no question that we can make that happen.
What will people do with all this computing power? They will find ways to change the world.