Network functions virtualization (NFV) continues to gain attention as a paradigm shift in the way telecommunications services are deployed and managed. By separating network function from traditional middle boxes, NFV is expected to lead to reduced capital expenditure and operating expenditure, and to more agile services. However, one of the main challenges to achieving these objectives is how physical resources can be efficiently, autonomously, and dynamically allocated to virtualized network function (VNF) whose resource requirements ebb and flow. To address this problem, we have developed a graph neural network-based algorithm which exploits VNF forwarding graph topology information to predict future resource requirements for each VNF component (VNFC). The topology information of each VNFC is derived from combining its past resource utilization as well as the modelled effect on the same from VNFCs in its neighbourhood. This project seeks to investigate the commercial viability of such a solution by developing a prototype demonstrator capable of handling commercial grade system performance, as the current algorithm has been validated to a limited degree via a simplistic virtualised service function chain.
The project advocates a dynamic and scalable management of resources to suppose virtualised Service Function Chains. Our approach automates resource management in NFV through a machine learning approach, i.e., graph neural network (GNN), which exploits the VNF topology information & its past resource utilization to predict future resource requirements. This is clearly a progress beyond the state of the art academically.
With regards to commercial activities, two of the core objectives of the project are to:
- Validate the problem space against target customers. The project aims to validate that our potential customers have the problem we think they have and that they’ll be interested in our solution.
- Develop an initial business case for further development. It will validate our initial assumptions through customer engagement and interviews
The project focused on setting up a sandbox commercial installation of the openstack Platform within the TSSG Datacentre testbed to deploy and evaluate the AutoScale. This testbed is composed of Software Defined Network devices and a suite of Virtual Network Functions. The testbed is configured to an identified realistic Service Chain Function, i.e., Clearwater IMS system. The project investigates the performance of Autoscale with respect to its measured impact on resource utilization of the Service Function Chains using the testbed.
The project is in process to validate the system to a commercial sandbox setting with an industry partner to validate the system in a closer to market environment, thereby aims to make the autoscale standard compliant. The project plan to carry trial with Dell EMC labs in Dublin, however other partners will also be sought, including Nokia Bell Labs, ERICSSON, HUAWEI, CISCO and Intel. All of which are members of the SFI CONNECT Research Centre.
The project addresses the following technical objectives:
- Develop a commercial prototype system integrated into the Openstack. The demonstrator shall be ready by the end of May this year.
- Validate the usage of GNN algorithm in capturing realistic resource requirements of Virtual Network Functions involved in realistic settings.
- Validate the system in an industry relevant environment in collaboration with an Industry partner (TRL 5)
- Determine a viable route to market along with a credible business case to commercialise.
The project has successfully developed a commercial prototype system integrated into the Openstack. The demonstrator shall be ready by the end of May this year. The demonstrator shall validate the usage of GNN algorithm in capturing realistic resource requirements of Virtual Network Functions involved in realistic settings
After validation of demonstrator, the project plans to offer the system as a Software as a Service network function to Telecommunication Service Providers directly as an add on to existing systems. For this course, we are making our system as standard compliant to integrate into the openstack platform.
The project has also initiated validation of the problem space against target customers with the assistance of pre-commercialization training offered under spark pre-accelerator program and National Digital Research Centre accelerator program funded by SFI. Based on this training, the project is developing an initial business case for further development.
Alan Davy, firstname.lastname@example.org
Shagufta Henna, email@example.com