Experimental Results

Best Case Numbers

Synthetic load:
-N requests for same file
=> best case

Measure Bandwidth
-BW = N * bytes(F)/time
-File size 0-200kb
=> vary work per request

Observations:
-all exhibit similar results
-SPED has best performance
– Flash AMPED extra check for memory presence
– Zeus has anomaly

When data is in cache:
– SPED >> Amped Flash
– SPED and AMPED Flash >> HT/MP
With disk-bound workload
– AMPED FLASH >> SPED