Integrated Analytics - Enabling Virtual SBCs in the Cloud

I am back again, with another topic in my continuing series about the key attributes required for successful deployment of virtual, cloud-based SBCs. This time I am going to talk about integrated analytics and the importance of automating management in a virtual, cloud deployment.

Data creation and collection in a traditional, hardware-based SBC deployment stays local. An SBC creates event logs, CDRs, trace records, or performance and utilization statistics, and stores them on a local disk. While this information can be forwarded to a central element management system, the value of this data is primarily applicable to each individual SBC.

Unfortunately, in a virtual, cloud deployment, this local approach to data creation and collection just does not work. SBC VNF instances are dynamically created and retired, calls or sessions are allocated to different SBC VNF instances based on load balancing, and storage for SBC VNF instances is typically ephemeral and may not persist after VM termination. To address these issues, we needed to adopt a cloud-optimized solution for data creation, collection, and analysis.

Our solution has a data agent running with each VNF to capture key information and to have that data forwarded to a centralized data warehouse for storage and analytics. Our data warehouse is a write-optimized, cloud-based, storage solution that has been designed for both scale and fault tolerance.

A few examples of the metrics we will capture for VNFs are: memory usage percentage; average and peak CPU usage; system congestion level; average call rate; call arrival rate; and license utilization. Analytics of this data can be based on individual metric, an aggregation of metrics, or based on derivations of them, using mathematical and statistical functions.

Two use cases described below will highlight the value of integrated analytics to the success in a virtual, cloud-based deployment. The first is for automated VNF management, the second is VNF monitoring and troubleshooting.

Automated VNF management begins with a feedback loop of VNF analysis back into a VNF Manager or Service Orchestration system. With this feedback, it is possible to answer fundamental questions for VNF management: are new VNF instances required; should existing VNFs be retired; do I need to modify VNF attributes; or am I properly scaled for the traffic demand? Most importantly, with answers to these questions, the goal of on-demand VNF management can be achieved. And with automated on-demand VNF management, a service provider can optimize the elasticity and dynamic scaling that is enabled by a cloud deployment.

VNF monitoring and troubleshooting will also use the construct of automated feedback to raise alerts and alarms, either proactively or reactively based on data analytics. But the difference for monitoring and troubleshooting is that it often will encompass additional information, not typically described as traffic or utilization data. For example, the data agent running with the VNF will also capture and forward logs, telemetry information and call detail data.

With this information, Sonus provides the framework for SIP ladder diagrams, and the ability to look at complete call detail records. Furthermore, by having a centralized repository of this information it will no longer be necessary to log in and look at individual VNFs in order to ensure proper performance and functionality.

For Sonus, it is our goal to provide a robust, flexible, and an inherently indispensable management solution that is optimized for the virtual, cloud-based deployment model. By doing this, we ensure our service provider customer’s success in deploying virtual, cloud-based SBCs.