Prometheus: A flexible toolkit for the experimentation with virtualized infrastructures

Abstract

Summary A typical problem that arises when devising a novel resource management strategy for virtualized infrastructures is how to experimentally assess its ability of achieving its design goals and whether it advances the state of the art. Among the available options, the use of a physical testbed is usually considered to be the most appropriate because of its high degree of accuracy. Unfortunately, however, physical testbeds are characterized by a limited controllability of the experimental conditions. Moreover, their implementation is usually a complex and time‐consuming task that requires the integration of many software components that need to interact among them in non‐trivial ways. In this paper, we address the aforementioned issues by proposing Prometheus, a toolkit specifically designed to support the configuration, deployment, and use of physical testbeds suitable to perform experimental studies of resource management strategies and that provides a high degree of controllability and low implementation costs. We discuss the design, implementation, and use of Prometheus, and we show how it can be used in practice to configure and deploy a physical testbed by providing various examples of experimental activities that can be carried out by means of it.

Profit‐aware coalition formation in fog computing providers: A game‐theoretic approach

“…First, a profiling experiment is carried out by running VM i on a reference fog node x (eg, using the methodology described in 13–14,22 ) in which U x , i is progressively increased until its measured request processing rate reaches the value μ i , and the corresponding value of allocated physical CPU capacity is recorded.…”

“…For example, the co-location facility of geographic area j in Figure 1 hosts fog nodes that belong to FIPs x and y. Fog nodes rely on a virtualization platform to run the VMs where the various applications are encapsulated together with their software stack (see below). We assume that this platform provides suitable dynamic resource allocation mechanisms (see, for instance, other works 13,14,[20][21][22] able to partition the physical capacity of the fog node across the VMs it runs.…”