Sun’s HPC Watercooler points to an article in GridToday by Univa UD’s Rich Wellner.
[Rich] questions the merits of measuring utilization rates in cluster or grid. Instead, says Wellner, a factor like latency has far more impact on an organization’s effectiveness and core business missions.
You can find the original article here. GT also has a piece in which GT editor Derrick Harris analyzes Rich’s viewpoint and identifies different classes of users that place different values on utilization. Derrick’s piece is here.
Know your audience. The good and bad of “grid” is how many concerns it can address. Beyond the parlor game HPC people play “What do you mean by Grid and Clusters”, can we come up with a common prism / framework.
My start. What’s the balance of your priorities:
1. Utilization – CAPEX
2. Time to Solution – Business View
3. Automation – OPEX IT View
4. Flexibility to meet business need / Oppt Cost – App View
5. Application Performance – User View
David – interesting point.
When it comes to measurement, we’ve seen a couple user segments, that tend to have very different trade-offs depending on what they’re trying to do. They tend to have very different answers to this question:
Which would you prefer:
1. The ability to run 5X faster than the desktop after 1 day of coding
2. The ability to run 25X faster than the desktop after 1 week of coding
3. The ability to run 125X faster than the desktop after 3 weeks of coding