The Inefficiency

Can we derive the efficieny of the system from some basic parameters such as memory bandwidth. etc.




...."NCSA had been running Warewulf on a test cluster, but was eager to get a first-hand look at how it worked on a large production cluster. The results were exceptional. NCSA witnessed a significant increase in performance with Warewulf on LBNL's cluster. A final Linpack benchmark run on the cluster produced 1.5 teraflops, indicating an efficiency of 83.43 percent. NCSA is now planning to use Warewulf to administer its T2 cluster and is hoping for a similar improvement in performance."

I mean WTF. We should be able to quantify this stuff.