Joint Power Scaling of Processing Resources and Consolidation of Virtual Network Functions

Abstract

With the advent of network 'softwarization', Network Functions Virtualization (NFV) is foreseen to provide flexibility and programmability levels that would essentially help in coping with tomorrow's demands. However, energy efficiency and the resulting complexity of network/service management pose serious sustainability and scalability issues that may hinder NFV's advantages. This paper considers these aspects in the context of datacenter networks. We propose an energy-Aware resource allocation scheme to manage virtual machines, dedicated to perform certain (virtualized) network functions, among a pool of energy-Tunable physical resources (processors/cores). We use online measurements to periodically estimate some statistical features of the offered workloads by considering a fairly general renewal model that captures traffic bustiness and hardware operational settings. Then, resources are dynamically managed by jointly performing power scaling and in-server consolidation according to the actual workload variations. The average power consumption generated by this strategy is evaluated and compared with that of a classical bin-packing consolidation, over processors running always with the highest-performance configuration. Results show that the proposed approach can reduce the average power consumption of the datacenter by up to 10%, suggesting a considerable amount of annual savings

A Model-Based Approach Towards Real-Time Analytics in NFV Infrastructures

“…Most of the works in the literature regard estimating the system or queueing latencies towards efficient (QoS-aware) network service provisioning, considering networks of queues to model interactions among service chain/virtual system/VNF components, with each component modeled as a (unique) queueing system [21]- [24]. Delving deeper into the infrastructure level, [25] takes into account the impact of interrupt coalescing (IC) in the Network Interface Card (NIC) on VNF performance (in terms of latency and packet loss), while [26] considers the burstiness of the VNF workloads (both incoming and aggregated) in the power modeling.…”

“…In this work, we adopt the queueing model considered in [26], analytically characterizing the different aspects of the system; starting from there, a model-based analytics approach that uses -and adds value to -available PMCs is proposed towards real-time VNF workload profiling, as well as power and latency estimation.…”

“…We adopt the power consumption model proposed in [26] for an energy-aware core running VMs. The model works according to a renewal process, where the idle (I) and delay busy (SET + B) periods constitute independent and identically distributed (iid) "cycles" (R), as illustrated in Fig.…”

Modeling Performance and Energy Efficiency of Virtualized Flexible Networks

“…With these consideration in mind, we can turn back to the modelling aspects, and to the related control strategies that can be devised to jointly optimise performance and energy consumption. In this respect, as a possible example, we briefly summarise here the approach that has been taken in [37]. The scenario addressed is represented in Fig.…”

“…The scheme includes energy-and performanceaware workload classification rules that define the most energy efficient configuration to be applied to the serving core. Evaluations performed in [37] show that, despite VMs' workload variations, the total system workload is kept relatively stable, at approximately 18% below SLA specifications, and the policy applying both in-server consolidation and power scaling can gain about 10% with respect to baseline scenarios without the same capabilities. In a scaled-down datacenter example with 500 servers (with 2 octa-core processors each) and 10,000 VMs, at the average European Union electricity prices for industrial consumers of 0.12 €/kWh during the second half of 2014, this can turn to an annual saving of approximately 19,000 €.…”