Deployment Approach for M-CORD – An Open Source Reference Solution for 5G Wireless networks
As we all know, Central Office Architected as Data Center (CORD) is a general-purpose service delivery platform. It can be configured to host a set of services in support of residential (known as Residential CORD), enterprise (known as Enterprise CORD), mobile technologies (known as M-CORD), cross-cutting capabilities (Analytics for CORD) and emerging edge applications (IoT, gaming, VR).
M-CORD, the open reference solution for service-driven 5G architecture, includes the release of end-to-end open source slicing from programmable Radio Access Network (RAN) to disaggregated and virtualized Evolved Packet Core (EPC) and M-CORD Mini1. These open source building blocks will facilitate the transition from the traditional Central Office to a virtualized data center. In this blog, we will discuss and compare two different approaches for M-CORD deployment: Top-down vs. bottom-up approach and put forth the pros and cons of each.
Top-down approach for M-CORD Deployment
M-CORD uses XOS as orchestration framework and sees traction towards using ONAP.
XOS brings Everything-as-a-Service (XaaS) approach to CORD. It’s a service orchestration layer built on top of OpenStack and ONOS that manages scalable services running in a Central Office Rearchitected as a Datacenter (CORD). XOS unifies the management of SDNbased control applications, NFVbased data plane functions, and traditional cloud services.
Open Network Automation Platform (ONAP) provides a comprehensive platform for real-time, policy-driven orchestration, and automation of physical and virtual network functions which enables IT provides and developers to rapidly automate the new services. Both take a top-down approach where the Orchestration framework is chosen, which offers lots of in built-in and production deployment readiness features, which include redundancy, infra monitoring, automated deployment, scalability, etc. The user needs to build the components/services as per the requirement and ready to deploy.
Overall it looks promising with ready to use in-build features. However, when it comes down to deployment, we might see performance challenges in the components (specifically to data plane NFV) to be deployed using framework w. r. F networking, which is the core requirement in the Telecom world. So, if we hit the issue it will be at a very late stage and also, might end up with more time investigating it and bottoming up the issues and which might lead to critical changes to the heart of the entire framework.
Bottom-up approach for M-CORD Deployment
Considering the cons in Top down approach, we decided to explore the bottom-up approach where we focus on system performance as the foremost acceptance criteria. While researching the roadblocks in system performance, we discovered that we might require the system level architecture changes or networking changes. We first carried out performance analysis for DPDK based Data Plane node in bare metal, and then with KVM virtualization. We also tried implementing micro service platform with the bottom-up approach using Docker instead of Kubernetes/Mesos, while choosing the networking option, which gives us optimum performance. This approach helped us understand the performance bottlenecks at the lower level. However, now, we need to develop the deployment framework from the grounds up.
After carefully analyzing both the approaches, we realized that they complement each other and are working towards converging them to develop an optimal solution. We at Great Software Laboratory are weighing the benefits provided by the built-in open-source M-CORD framework vs. the performance bottlenecks at ground level.