Check out the new USENIX Web site. next up previous
Next: 4.1 Macro-benchmarks Up: Trading Capacity for Performance Array Previous: 3.5 Validating the Integrated

   
4. Experimental Results

In this section, we evaluate the performance of the prototype MimdRAID under two sets of experiments. The first set of experiments is based on playing real-world file system and transaction processing traces on the MimdRAID simulator. The second set of experiments is based on running on the prototype itself a synthetic workload generator that is designed to stress it in ways beyond what is possible with the traces. The purposes of the experiments are to: 1) validate the models of Section 2, 2) show the effectiveness and importance of workload-driven configuration, and 3) demonstrate the use of scaling the number of disks as a cost-effective means of improving performance for certain workloads.



 

Xiang Yu
2000-09-11