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

Abstract

Network Functions Virtualization (NFV) has recently gained much popularity in the research scene for the flexibility and programmability that it will bring with the software implementation of network functions on commercial offthe-shelf (COTS) hardware. To substantiate its roll out, a number of issues (e.g., COTS' inherent performance and energy efficiency, virtualization overhead, etc.) must be addressed, in a scalable and sustainable manner. Numerous works in the scientific literature manifest the strong correlation of network key performance indicators (KPIs) with the burstiness of the traffic. This paper proposes a novel model-based analytics approach for profiling virtualized network function (VNF) workloads, towards real-time estimation of network KPIs (specifically, power and latency), based on an M X /G/1/SET queueing model that captures both the workload burstiness and system setup times (caused by interrupt coalescing and power management actions). Experimental results show good estimation accuracies for both VNF workload profiling and network KPI estimation, with respect to the input traffic and actual measurements, respectively. This demonstrates that the proposed approach can be a powerful tool for scalable and sustainable network/service management and orchestration.

A Categorization of Power Measurement and Power Consumption Models in the Telco Cloud

“…A7: Controlled resource provision [64,90,95,97]. Provision of resources may change during the running of an experiment (rather than before it starts).…”

“…Characterizations may need to be sharper. For example, since packet network traffic arrival is known to often have the properties of a Batch Markov Arrival (stochastic) Process (BMAP), this is an operating constraint adopted in the approach of several works studying power efficiency of network functions [64,95,105,111].…”

“…4 average network transmission rate, as a predictor of host system power [81] 5 statistics of a Batch Markov Arrival Process (BMAP) (packet traffic) as a means of prediction of power consumption by a VNF [95,105] Hardware is abstracted through measurement of power consumption at some operating point, or change in power consumption over some operating range. Examples follow.…”

“…5 In [95], the operating point is the power consumption when a VE running on a single processor core is switching packets at the maximum rate for a given performance state.…”

“…With one exception, none of the works in the above list uncovers the hood to peer at the processor's internals (to obtain predictors of power consumption). The exception is [95]; yet even in this case, the performance monitoring counters are not used as direct predictors of power consumption, but to obtain (a) the timing information necessary for a queueing model and (b) the operating state (ACPI (Advanced Configuration and Power Interface [113]) P-and C-states).…”