Allocating and Scheduling Measurement Applications

Improvements in containerization and orchestration have facilitated the management of distributed measurement platforms, helping rationalize networks and services monitoring, a vital but demanding activity. Nonetheless, orchestration in this domain, still requires optimization, which we address in this paper. More precisely, we invest our efforts in the optimization of allocating and scheduling network and service measurement applications, deployed in a pool of physical machines in the network. First, we formulate the problem into an Integer Linear Programming model and prove it to be NP-Hard. Then, we propose an approximate resolution based on the Ant Colony System (ACS) metaheuristic. Finally, we introduce an open-source distributed measurement platform implementing our algorithm. Results show that the ACS algorithm provides efficient orchestration solutions that outperform those provided by a Branch and Bound algorithm and a First Fit Decreasing algorithm.