In a Telco environment, developers often leverage cloud-offloading to improve application or VNF performance, and as more workloads migrate to Cloud, hardware acceleration has already become a necessity rather than an interesting option. Cyborg provides the ability to manage accelerators in an OpenStack framework.

In this talk, we will provide reasons for needing Cyborg; including the inconvenience of managing different kinds of hardware workloads that are beyond the capabilities of core OpenStack projects like Nova or Neutron. We will then talk about how we utilized Cyborg to manage accelerator resources like Netronome Network Card, Xlinux FPGA Card and Intel's QAT card. This particular section will cover what we have done with Cyborg API, Conductor, drivers, as well as Nova or Neutron in order to follow and enhance Cyborg’s capabilities.

We will then conclude by providing our experience using Cyborg and potential pitfalls that the operators need to be mindful off.

Search Result

Schedule Matches

Speaker Matches