A Queuing Based Dynamic Auto Scaling Algorithm for the LTE EPC Control Plane

Abstract

This is the accepted version of the original article published by IEEE.Abstract-Network softwarization paradigm, whose main enabler is Network Functions Virtualization (NFV), facilitates the automation of the management operations and orchestration of the future networks, thus reducing the operational expenditures of the network. The envisioned management practices include the introduction of automation in the scaling of network services. This may enable operators to handle workload fluctuations to keep the desired performance with great agility and reduced costs. This procedure introduces a non-negligible delay in allocating or releasing virtual resources. Therefore, waiting until the system is overloaded or underutilized so as to scale resources up or down could negatively impact user Quality of Experience, or lead to inefficient resource utilization. In this vein, this paper proposes a novel and agile Dynamic Auto Scaling Algorithm for the Long Term Evolution (LTE) virtualized Evolved Packet Core (vEPC) Control Plane (CP). The resources dimensioning stage of the algorithm is based on an original queuing model for the LTE CP. To model the LTE CP, we use an open network of G/G/m queues. We also provide expressions to derive the steady state transition probabilities of the queuing network. Finally, we validate the proper operation of our solution using accurate simulation tools.

Performance Modeling of Softwarized Network Services Based on Queuing Theory With Experimental Validation

“…An initial version of the performance model proposed in this paper was described in our previous work [14]. After, it has been applied to the planning and DRP for specific scenarios in [3] and [6], [7], respectively. All those previous works have reported satisfactory simulation results on the accuracy and usefulness of the model.…”

“…For instance, given an overall delay budget of the system, it is possible to define algorithms that rely on QN models to optimally distribute the delay budget among the different VNFCs. This approach leads to resources saving, as shown in [6]. Given the present state of the art, only those QNs that admit a product-form solution, i. e., the joint probability of the QN states is a product of the probabilities of the states in individual queuing nodes, can be analyzed precisely [15], [16].…”

On Monolithic and Microservice Deployment of Network Functions

“…The authors in [26] have proposed a three-tier queuing network for modelling VNF chains. In [27], a similar model is used to represent a virtualized Evolved Packet Core control plane which the authors use to investigate the performance impact of scaling procedures and propose an automatic scaling algorithm. The proposed model is validated through simulation.…”

Closed-Form Expression for the Resources Dimensioning of Softwarized Network Services

“…There are several heuristics proposed in the literature to tackle the Resources Dimensioning (RD) problem of softwarized network services [11], [13], [18], [20]. In [11], the authors formulate and propose a heuristic to solve the joint optimization problem for the Service Function Chain (SFC) routing and VNF instance dimensioning.…”

“…Then, the authors propose a heuristic method to address the issue. In [18], [20], the authors propose a heuristic to carry out the joint RD of the Control Plane (CP) entities of a virtualized EPC. The heuristic requires an auxiliary methodology to predict the vEPC response time for a given setup.…”