Workload generation. We use httperf [26] to generate synthetic workloads for RUBiS. We gen- erate workloads for specific durations with a required number of user sessions per second. A user session emulates a visitor that browses items up for auc- tion in specific categories and geographical regions and also bids on items up for auction. In a first cycle, every five minutes, we increment the load level by 6, from load level 6 up to load level 108, and then we decrement the load level by 6 from load level 108 down to load level 6. In a second cycle, we increment the load level by 6, from load level 6 up to load level 60, and then we decrement the load level by 6 from load level 60 down to load 120 108 96 Load level (new sessions/s) 84 72 60 48 36 24 12 0 0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 Time (minutes) Figure 4: Workload generation profile for all experiments.
Workload generation. I developed a shell script to generate synthetic workload for the sample Web application. The shell script contains the following httperf command: httperf --hog --timeout=100 --server=10.0.0.1 --port=80 --wsesslog=10000000,0,input.txt --rate $1 -v The parameter –rate is used to define the number of sessions created per second. Usually, Web applications users make requests interactively and take some time during requests, known as user think time. The parameter –wsesslog defines the total number of sessions to generate, user think time, and input file name. The input file allows specification of a sequence of requests, the request method, and user think time per burst. The shell script allows control of the duration of the workload. My input.txt file contains ten requests of four bursts and each burst contains some think time in seconds. This script allows generation of workload
