Methods for efficiently processing sets of interdependent tasks in a distributed environment

Due to the emergence of the Internet of Things, smart buildings and smart cities, increasingly large amounts of data received from sensors and devices have to be processed in near real-time. All the information is processed in highly computational distributed environments and is made available on demand or is used by intelligent systems to control various devices. In some cases, the received data is relevant and can be processed only after other pieces of information have been handled. This paper analyses and proposes methods for efficiently processing all the data in the said environment. The problem is translated into the problem of mapping jobs (sets of interdependent tasks) to processing units by considering that a job represents the processing of multiple dependent pieces of data. In order to solve this mapping problem, six methods that overcome the task dependency issues from the Min-Min and Max-Min heuristics are proposed and compared using three performance metrics: makespan, load balance and processing unit idle time. The tests have showed that the proposed methods are efficient in solving the mapping problem but the degree of efficiency depends on the performance metrics that the system needs to improve.