Quality of service trade-offs between Central Data Centers and nano data centers

Nano data centers are one of the latest trends in cloud computing aiming towards distributing the computing power of massive data centers among the clients in order to overcome setup and maintenance costs. The distribution process is done over the already present computing elements in client houses such as tv receivers, wireless modems, etc. In this paper we investigate the feasibility of using nano data centers instead of conventional data centers containing accumulated computing power. We try to draw the lines that may affect the decision of nano data center approach considering important parameters in cloud computing such as memory capacity, diversity of user traffic and computing costs. We also investigate the thresholds for these parameters to find out the conditions that make more sense to set up nano data centers as the best replacement of Central Data Centers. We use a CloudSim based simulator, namely CloudAnalyst, for Data Center performance experiments in Java. Our results show that 1 gigabyte memory capacity can be seen as a threshold for response time improvement of nano data centers. For nano data centers with more memory capacity there will not be any improvement in response times that leverages the performance cost. We also combine the results of response time and performance cost to provide a similar threshold.