NETbalance: Reducing the Runtime of Network Emulation Using Live Migration

Abstract

Network emulation is an efficient method for evaluating distributed applications and communication protocols by combining the benefits of real world experiments and network simulation. The process of network emulation involves the execution of connected instances of the software under test (called virtual nodes) in a controlled environment. In previous work, we introduced an approach to minimize the runtime of network emulation experiments based on prior known average resource requirements of virtual nodes.In this paper, we introduce NETbalance, a novel approach to runtime reduction for experiments with unknown or varying resource requirements. NETbalance migrates virtual nodes during an experiment to distribute the load evenly across the physical nodes, avoiding overloaded nodes and exploiting the idle resources on underloaded nodes for speeding up the experiment execution. We make the following contributions: First, we present an emulation architecture for efficiently supporting live migration of virtual nodes. Second, we propose a cost model for determining the runtime reduction achieved through the migration. Third, we introduce an algorithm for calculating placements that minimize the experiment runtime. Our evaluations of the NETbalance prototype show, that it is able to reduce the experiment runtime by up to 70%.

A Virtual Time System for Linux-container-based Emulation of Software-defined Networks

“…It was first introduced by Gupta et al [25], and has been adopted to various types of virtualization techniques and integrated with a handful of network emulators. Examples include DieCast [24], SVEET [19], NETbalance [22], TimeJails [23,35] and TimeKeeper [32]. The second approach focuses on synchronized virtual time by modifying the hypervisor scheduling mechanism.…”

Integrated simulation and emulation using adaptive time dilation

“…NETplace [23] and NETbalance [24] use epoch-based virtual time [22] for implementing a dynamic time dilation. While the dynamic time dilation in [22] uses a threshold-based load control mechanism, our TDF controller maintains system loads at a target level.…”

High-performance emulation of heterogeneous systems using adaptive time dilation

“…For example, PlanetLab (Chun et al, 2003) is based on OSlevel virtualization that uses Linux-VServer, 7 which creates many independent containers under a common Linux kernel. OpenVZ, which also adopts OS-level virtualization, was employed to create virtual nodes in EMULAB, DeterLab, NETplace (Grau et al, 2010), NETbalance (Grau et al, 2011), and the other network emulation studies (Grau et al, 2008(Grau et al, , 2009Zheng and Nicol, 2011). Lightweight virtual nodes thus have been widely used for scalable network emulations, but have the limitation that they all support a single type of OS.…”

“…Both approaches, as introduced in Weinga¨rtner et al (2008Weinga¨rtner et al ( , 2011 and Gupta et al (2005Gupta et al ( , 2008, are static in that the relative ratio between real and virtual time is fixed for the life of VMs. On the other hand, NETplace (Grau et al, 2010) and NETbalance (Grau et al, 2011) use epoch-based virtual time (Grau et al, 2009) for implementing a dynamic TDF. By setting TDF to infinity, Grau et al (2009) effectively suspend the advance of virtual time; however, since the VMs continue to run, the results of the emulation may be unrealistic (as the VMs perceive infinite CPU speed at infinite TDF).…”