skip to main content


Title: A Reinforcement Learning-Based Admission Control Strategy for Elastic Network Slices
This paper addresses the problem of admission control for elastic network slices that may dynamically adjust provisioned bandwidth levels over time. When admitting new slice requests, sufficient spare capacity must be reserved to allow existing elastic slices to dynamically increase their bandwidth allocation when needed. We demonstrate a lightweight deep Reinforcement Learning (RL) model to intelligently make ad-mission control decisions for elastic slice requests and inelastic slice requests. This model achieves higher revenue and higher acceptance rates compared to traditional heuristic methods. Due to the lightness of this model, it can be deployed without GPUs. We can also use a relatively small amount of data to train the model and to achieve stable performance. Also, we introduce a Recurrent Neural Network to encode the variable-size environment and train the encoder with the RL model together.  more » « less
Award ID(s):
2008856
NSF-PAR ID:
10343822
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
2021 IEEE Global Communications Conference (GLOBECOM)
Page Range / eLocation ID:
01 to 06
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. This paper addresses a routing selection strategy for elastic network slices that dynamically adjust required resources over time. When admitting elastic initial slice requests, sufficient spare resources on the same path should be reserved to allow existing elastic slices to increase their bandwidth dynamically. We demonstrate a deep Reinforcement Learning (RL) model to intelligently make routing choice decisions for elastic slice requests and inelastic slice requests. This model achieves higher revenue and higher acceptance rates compared to traditional heuristic methods. Due to the lightness of this model, it can be deployed in an embedded system. We can also use a relatively small amount of data to train the model and achieve stable performance. Also, we introduce a Recurrent Neural Network to auto-encode the variable-size environment and train the encoder together with the RL model. 
    more » « less
  2. A key dimension of reproducibility in testbeds is stable performance that scales in regular and predictable ways in accordance with declarative specifications for virtual resources. We contend that reproducibility is crucial for elastic performance control in live experiments, in which testbed tenants (slices) provide services for real user traffic that varies over time. This paper gives an overview of ExoPlex, a framework for deploying network service providers (NSPs) as a basis for live inter-domain networking experiments on the ExoGENI testbed. As a motivating example, we show how to use ExoPlex to implement a virtual software-defined exchange (vSDX) as a tenant NSP. The vSDX implements security-managed interconnection of customer IP networks that peer with it via direct L2 links stitched dynamically into its slice. An elastic controller outside of the vSDX slice provisions network links and computing capacity for a scalable monitoring fabric within the tenant vSDX slice. The vSDX checks compliance of traffic flows with customer-specified interconnection policies, and blocks traffic from senders that trigger configured rules for intrusion detection in Bro security monitors. We present initial results showing the effect of resource provisioning on Bro performance within the vSDX. 
    more » « less
  3. Availability is a key service metric when deploying service function chains (SFCs) over network slices in 5G networks. We study the problem of determining the composition of a slice for a service function chain and the mapping of the slice to the physical transport network in a way that guarantees availability of the SFC while minimizing cost. To improve the availability, we design a slice that provides multiple paths (possibly with non-disjoint routing over the physical infrastructure) for hosting SFCs, and we determine the appropriate dimensioning of bandwidth on each path. Our simulation results show the effectiveness of our approach in terms of the cost of establishing the SFC and the SFC acceptance ratio.

     
    more » « less
  4. null (Ed.)
    Efficient provisioning of 5G network slices is a major challenge for 5G network slicing technology. Previous slice provisioning methods have only considered network resource attributes and ignored network topology attributes. These methods may result in a decrease in the slice acceptance ratio and the slice provisioning revenue. To address these issues, we propose a two-stage heuristic slice provisioning algorithm, called RT-CSP, for the 5G core network by jointly considering network resource attributes and topology attributes in this paper. The first stage of our method is called the slice node provisioning stage, in which we propose an approach to scoring and ranking nodes using network resource attributes (i.e., CPU capacity and bandwidth) and topology attributes (i.e., degree centrality and closeness centrality). Slice nodes are then provisioned according to the node ranking results. In the second stage, called the slice link provisioning stage, the k-shortest path algorithm is implemented to provision slice links. To further improve the performance of RT-CSP, we propose RT-CSP+, which uses our designed strategy, called minMaxBWUtilHops, to select the best physical path to host the slice link. The strategy minimizes the product of the maximum link bandwidth utilization of the candidate physical path and the number of hops in it to avoid creating bottlenecks in the physical path and reduce the bandwidth cost. Using extensive simulations, we compared our results with those of the state-of-the-art algorithms. The experimental results show that our algorithms increase slice acceptance ratio and improve the provisioning revenue-to-cost ratio. 
    more » « less
  5. The high reliability required by many future-generation network services can be enforced by proper resource assignments by means of logical partitions, i.e., network slices, applied in optical metro-aggregation networks. Different strategies can be applied to deploy the virtual network functions (VNFs) composing the slices over physical nodes, while providing different levels of resource isolation (among slices) and protection against failures, based on several available techniques. Considering that, in optical metro-aggregation networks, protection can be ensured at different layers, and the slice protection with traffic grooming calls for evolved multilayer protection approaches. In this paper, we investigate the problem of reliable slicing with protection at the lightpath layer for different levels of slice isolation and different VNF deployment strategies. We model the problem through an integer linear program (ILP), and we devise a heuristic for joint optimization of VNF placement and ligthpath selection. The heuristic maps nodes and links over the physical network in a coordinated manner and provides an effective placement of radio access network functions and the routing and wavelength assignment for the optical layer. The effectiveness of the proposed heuristic is validated by comparison with the optimal solution provided by the ILP. Our illustrative numerical results compare the impact of different levels of isolation, showing that higher levels of network and VNF isolation are characterized by higher costs in terms of optical and computation resources.

     
    more » « less