PISCO: A smart simulator to deploy energy saving methods in microservices based networks
Abstract
Nowadays, many researchers work to identify microservices-based application deployments and scheduling solutions to save energy without decreasing functional QoS. In this work, we present PISCO: A simulator that allows facing this challenge in a simple and efficient way, enabling its users to focus uniquely on microservices deployment/scheduling algorithms and its hardware/software repercussions (load vs. energy consumption) without worrying about low-level network configurations or operating system issues. PISCO is able to deploy and schedule (move, duplicate, start/stop) microservices and their dependencies on various devices with software and hardware heterogeneity (CPU, bandwidth, RAM, Battery, etc.), taking into account various scheduling heuristics algorithms: centralized vs noncentralized. To do this, PISCO allows deploying custom network topologies based on client-server schemes or p2p distributions, where devices can (dis)appear, turn on/off obeying random circumstances or user strategies. Finally, the simulator performs relevant operations such as QoS definition, resource monitoring, calculation of energy saved and consumption tracking (at device and network level). We tested some ideas based on our previous work "Kaligreen" to demonstrate the effectiveness of PISCO.
Domains
Computer Science [cs]
Origin : Files produced by the author(s)