Fig 1.
Architecture for the approach considering the new client entity.
This figure shows the architecture of the simulation organized in layers, considering the new client entity.
Fig 2.
Transition MAP Diagram–Adapted from [20].
Example of a two-state MAP process (Normal State and Burst State) to control the requests arrival rate in the system.
Fig 3.
Pseudo Code–defining the parameters of the MAP.
Script for fitting the parameters of a MAP to model bursts in the arrival process, considering the main measures as average, I and SCV of the MAP.
Table 1.
Sets of Experiments.
Fig 4.
Average response time in seconds for different ways of submission.
This graph shows the way that the requests arrive to be executed in the cloud influences in the average response time.
Fig 5.
Average system time in seconds for different ways of submission.
Each set of three bars considers a way of request submission, for both individual requests or in groups. Each way of submission has three bars, representing parts of the response time.
Fig 6.
Average Throughput in request/second for different ways of submission.
The experiment that consider dependent request group mode presents lowest average throughput, mainly influenced by the synchronization between requests of the group.
Fig 7.
Samples of response times for the different modes obtained of submission.
The graphics show the different behavior between client send modes considering the response time and the requests ID during the simulation time.
Fig 8.
Samples of the number of arrivals of requests for the different modes of submission.
The graphics show the number of requests arriving monitored at time intervals of 7 seconds in the simulation.
Fig 9.
Percentage of resource usage for the different modes of submission.
The graphics show samples of the percentage of resource usage, monitored at time intervals of 1 second in the simulation.
Fig 10.
Response Variables considering Bursty and Non-Bursty Workload with think time after response.
Experiments that consider workloads with and without burst and using think times beginning after receiving the response from the previous request (a) Average response time (b) Throughput (c) Average service time.
Fig 11.
Samples of the response time of requests considering Bursty and Non-Bursty Workload with think time after response.
Samples of the experiments where think times beginning after receiving the response from the previous request (a) without bursty, (b) with bursty.
Fig 12.
Samples of the number requests arrivals considering Bursty and Non-Bursty Workload with think time after response.
Samples of the experiments where think times begin after receiving the response from the previous request (a) without bursty, (b) with bursty.
Fig 13.
Samples of the resource usage percentage considering Bursty and Non-Bursty Workload with think time after response.
Samples of the experiments where think times begin after receiving the response from the previous request (a) without bursty, (b) with bursty.
Fig 14.
Response Variables considering Bursty and Non-Bursty Workload–think time after submission.
Experiments that consider workloads with and without burst and that consider the beginning of the think time immediately after submission of the previous request (a) Average response time (b) Average service time.
Fig 15.
Samples of the requests response time considering Bursty and Non-Bursty Workload with think time after submission.
Samples of the experiments that consider the beginning of the think time immediately after submission of the previous request (a) without bursty, (b) with bursty.
Fig 16.
Samples of the requests arrivals number considering Bursty and Non-Bursty Workload with think time after submission.
Samples of the experiments that consider the beginning of the think time immediately after submission of the previous request (a) without bursty, (b) with bursty.
Fig 17.
Samples of the resource percentage usage considering Bursty and Non-Bursty Workload with think time after submission.
Samples of the experiments that consider the beginning of the think time immediately after submission of the previous request (a) without bursty, (b) with bursty.
Fig 18.
Response Variables considering homogeneous and heterogeneous network latency.
Results of the experiments considering homogeneous and heterogeneous network latency between client and broker: (a) Average response time, (b) Average Throughput.
Fig 19.
Average System Time and Average Latency for experiments that consider homogeneous and heterogeneous network latency.
Results of the experiments considering homogeneous and heterogeneous network latency between client and broker.