Cooperative load balancing scheme for edge computing resources

Edge Computing, as a solution to leveraging computation capabilities at the edge of the network, is emerging. One key challenge for edge computing is offering its computing service with a low service blocking and low latency that otherwise translate to an inefficient deployment of a Edge computing system. Unlike cloud computing, the computing resources in edge computing are limited. One way to deal with this limitation is by enabling cooperation between data centers. In this paper we propose CooLoad, a cooperative scheme between two data centers installed at the edge of the network that exchanges each other generic computation requests when one of them is temporary overloaded. If a service request arrives at one data center with its request buffer full, then the request is forwarded to the other cooperating data center and served by that data center, consuming its CPU cycles. We define a mathematical model that captures the main aspects of the cooperation strategy and assesses the effectiveness of the scheme, showing how the service blocking probability and service delay can in principle be both reduced at the same time. We also propose a possible implementation of the cooperative scheme