%0 Thesis %1 info3-praktikum-2016-10 %A Züfle, Marwin %D 2016 %K i3thesis aquareyoum can %T Testbed Validation and Comparison with Simulation %X In this report the main components of a testbed for evaluating the performance of resource management strategies has been validated. Application-aware resource management is a very important task since high quality applications like video streaming, social networking, and online gaming need a more intelligent bandwidth allocation policy than best effort to reach the highest possible Quality of Experience for all end users. Several tests have been performed to validate the determination of the web page size, Emma, Dory, and TraSh. It has been shown that the web page size has to be determined by measuring the used bandwidth at the router using a tool like ifstat since HTTP and TCP headers are not considered by tools within the browser like Firebug. Additionally, measurements provided that the latency caused by sending the resource allocation commands from Dory to TraSh is very small and should not affect the resource management results. Furthermore, the different random number generator distributions have been examined. It has been shown that Emma typically generates too many request arrivals with too low interarrival time so that requests are rejected. The validation of TraSh has shown that the time needed to realize the resource allocation adaption is pretty short and that there is no statistically significant difference between this delay for different capacities. On top of that, several problems of TraSh have been detected. Finally, the download times of web browsing requests have been measured and compared to simulation results. It has been shown that the download times in the measurement are much higher than the ones of the simulation if there is no initial delay implemented. By implementing an initial delay in the simulation it was shown that the download times converge. However, the initial delay needs to be further investigated since the elapsed time is not reliable known. On top of that, the problems of TraSh need more examination and resolving. Finally, the resource management strategies need to be implemented in Dory in order to measure their impact on download times for web browsing and buffering ratio for video streaming.

@mastersthesis{info3-praktikum-2016-10, abstract = {In this report the main components of a testbed for evaluating the performance of resource management strategies has been validated. Application-aware resource management is a very important task since high quality applications like video streaming, social networking, and online gaming need a more intelligent bandwidth allocation policy than best effort to reach the highest possible Quality of Experience for all end users. Several tests have been performed to validate the determination of the web page size, Emma, Dory, and TraSh. It has been shown that the web page size has to be determined by measuring the used bandwidth at the router using a tool like ifstat since HTTP and TCP headers are not considered by tools within the browser like Firebug. Additionally, measurements provided that the latency caused by sending the resource allocation commands from Dory to TraSh is very small and should not affect the resource management results. Furthermore, the different random number generator distributions have been examined. It has been shown that Emma typically generates too many request arrivals with too low interarrival time so that requests are rejected. The validation of TraSh has shown that the time needed to realize the resource allocation adaption is pretty short and that there is no statistically significant difference between this delay for different capacities. On top of that, several problems of TraSh have been detected. Finally, the download times of web browsing requests have been measured and compared to simulation results. It has been shown that the download times in the measurement are much higher than the ones of the simulation if there is no initial delay implemented. By implementing an initial delay in the simulation it was shown that the download times converge. However, the initial delay needs to be further investigated since the elapsed time is not reliable known. On top of that, the problems of TraSh need more examination and resolving. Finally, the resource management strategies need to be implemented in Dory in order to measure their impact on download times for web browsing and buffering ratio for video streaming. }, added-at = {2017-01-03T17:11:26.000+0100}, author = {Züfle, Marwin}, biburl = {https://www.bibsonomy.org/bibtex/2208e4613c6e3a3685ae9e9675b1e229b/uniwue_info3}, interhash = {2c3d82f0f0488d6a661f04a6ece0cf7e}, intrahash = {208e4613c6e3a3685ae9e9675b1e229b}, keywords = {i3thesis aquareyoum can}, month = {11}, school = {University of Würzburg}, timestamp = {2022-03-14T00:08:36.000+0100}, title = {Testbed Validation and Comparison with Simulation}, year = 2016 }