Overview / Usage

**Everyone agrees Kubernetes is the future. But can it be now?This is 5G use case to let Kubernetes become reality for everyone now instead of waiting for the future to come. **

This solution uses an open source SDN controller to enable SFC (service function chaining) with existing PNF/VNF (physical and virtual). Then it adds application telemetry to provide intelligent bot detection as well network analysis to predict failures and heal the network automatically.

With this, existing telcos can offer addition services to their customers using existing equipment. There is no new CAPEX/OPEX and the technology is mature. Customers can feel confident to start deploying Kubernetes.

Without this, customers need to rely on Service Mesh, which is not yet commercially proven. And they need to invest in new solutions.

We believe this solution will stimulate the Kubernetes hockey-stick demands. Components used:

1. Open source SDN controller

2. Kubernetes

3. OpenStack

4. DPDK/VPP

5. Advanced application telemetry

6. Intelligent network security segmentation

7. Linux Foundation Acumos AI framework

​

Methodology / Approach

The solution consists of a service agent residing on every Kubernetes node and a controller. The diagram above shows the overall architecture and two SFC examples.

The service agent collects the telemetry information feeding to the controller. The controller has a gateway function, an analytics engine, and Acumos AI framework to business cases. The gateway function is based on DPDK or VPP for SFC. It can be offloaded to a separate server, but for the demo purpose, we’d put in on the same server.

The controller uses the telemetry collected by service agents and its own gateway functions then applies the intelligence to (1) detect any traffic anomaly and can either drop those packets or report the incidence; (2) predicts potential failures and use SDN control plane to steer the traffic to different paths; (3) if any unexpected fault happens, it heals the workloads by using Kubernetes CNI asking Kubernetes master to move the pods to the alternative network.

Controller can also be set at a supervisor mode where every change will need a confirmation from the designated system administrator. Controller will give recommendations based on its intelligence. It’s different from traditional controllers that only report alerts. Based on the recommendations, the system administrator can easily make a quick decision without needing to troubleshoot.

Technologies Used

Tungsten Fabric, NUC, OpenStack, Firewall, Load Balancer, BGP-EVPN, MPLS.

Collaborators

1 Result

* Sort

Most Popular Number of Activities Number of Projects Recently Joined

• Projects 1
• Followers 0

Ming Jiang

Follow

1 Result