34
3.5
Performance Analysis
In this study, we consider two performance metrics: average latency (response
time) and throughput. Latency is the time from the arrival of a request at an initial
server until completion of that request. Throughput is the number of requests completed
per second.
3.5.1
Response Time and Throughput Results
4
x 10
45
5.5
press_via
40
5
dcs
press_via
adaptive
35
dcs
4.5
dcs_block
30
adaptive
dcs_block
4
25
3.5
20
Throughput
3
15
Average Latency (ms)
2.5
10
5
2
0
1.5
10
8
6
4
2
10
9
8
7
6
5
4
3
k
k
(a) Latency
(b) Throughput
Fig. 3.6. Average Latency and Throughput of a 32 node Web Cluster
The latency and throughput results of PRESS over VIA (press via), coscheduled
PRESS with DCS (dcs), Adaptive model (adaptive) and DCS with blocking (blocking)
are plotted for a 32 node configuration in Figure 3.6. The results are measured as the
value of k decreases from 10 to 2.9, where k is the parameter of the Pareto distribution