Efficient network slicing is vital to deal with the highly variable and dynamic characteristics of network traffic generated by a varied range of applications. The problem is made more challenging with the advent of new technologies such as 5G and new architectures such as SDN and NFV. Network slicing addresses a challenging dynamic network resource allocation problem where a single network infrastructure is divided into (virtual) multiple slices to meet the demands of different users with varying requirements, the main challenges being -the traffic arrival characteristics and the job resource requirements (e.g., compute, memory and bandwidth resources) for each slice can be highly dynamic. Traditional model-based optimization or queueing theoretic modeling becomes intractable with the high reliability, and stringent bandwidth and latency requirements imposed by 5G technologies. In addition these approaches lack adaptivity in dynamic environments. We propose a deep reinforcement learning approach to address this dynamic coupled resource allocation problem. Model evaluation using both synthetic simulation data and real workload driven traces demonstrates that our deep reinforcement learning solution improves overall resource utilization, latency performance, and demands satisfied as compared to a baseline equal-slicing strategy.

In today's world, mobile phone penetration has reached a saturation point. As a result, subscriber churn has become an important issue for mobile operators as subscribers switch operators for a variety of reasons. Mobile operators typically employ churn prediction algorithms based on service usage metrics, network performance indicators, and traditional demographic information. A newly emerging technique is the use of social network analysis (SNA) to identify potential churners. Intuitively, a subscriber who is churning will have an impact on the churn propensity of his social circle. Call detail records are useful to understand the social connectivity of subscribers through call graphs but do not directly provide the strength of their relationship or have enough information to determine the diffusion of churn influence. In this paper, we present a way to address these challenges by developing a new churn prediction algorithm based on a social network analysis of the call graph. We provide a formulation that quantifies the strength of social ties between users based on multiple attributes and then apply an influence diffusion model over the call graph to determine the net accumulated influence from churners. We combine this influence and other social factors with more traditional metrics and apply machine‐learning methods to compute the propensity to churn for individual users. We evaluate the performance of our algorithm over a real data set and quantify the benefit of using SNA in churn prediction. © 2013 Alcatel‐Lucent.

No abstract

No abstract

Abstract-In this paper we take a closer look at the operation of software defined networking (SDN) in intra-domain networks. The focus is on the dependability issues related to interworking of SDN controllers, network OS (NOS), and forwarding in the data plane. Both the separation of the control and data planes, and the (virtually) centralized control processes, are challenging from a dependability perspective. In particular, consistency in operation and information is a challenge, both between the control and data planes, but also within a given plane. To ensure the necessary level of consistency there could be a conflict with the strict realtime requirements given by the per-flow operation of the SDN controller. A principle system model is introduced to discuss the consistency challenge, and to point out undesirable cyclic dependencies between functions that are necessary to configure and operate SDN. The separation of control processing and forwarding do also introduce structural vulnerabilities, which are exemplified.