|
To test PARDA and its robustness with mixed workloads, we ran very different workload patterns at the same time from our six hosts. Table 5 presents the uncontrolled case.
Next, we enable PARDA with = 40 ms, and assign shares in a ratio for hosts 1 through 6 respectively, plotted in Figure 10. Window sizes are differentiated between hosts with different shares. Hosts with more shares reach a window size of 32 (the upper bound, ) and remain there. Other hosts have window sizes close to 19. The average latency observed by the hosts remains close to , as shown in Figure 10(b). The throughput observed by hosts follows roughly the same pattern as window sizes, but is not always proportional because of array scheduling and block placement issues. We saw similar adaptation in window sizes and latency when we repeated this experiment using = 30 ms (plots omitted due to space constraints).