Netronome and Juniper are Accelerating Open Contrail vRouter at MWC

By Netronome | Feb 19, 2016

Here at Netronome, we are gearing up for an exciting week at Mobile World Congress, hosted in Barcelona on February 22-26. One activity that we are particularly excited about is the demonstration of Netronome’s Agilio CX SmartNIC and Agilio vRouter software inside the Juniper booth, which will highlight our ongoing collaboration with Juniper to provide an accelerated vRouter data plane for the Open Contrail system. In addition, a performance demonstration that boasts a vRouter data path capacity of 27Mpps for VNF service chaining, a 6X improvement over software-only vRouter, will be on display in Netronome’s hospitality suite (2N17MR, Hall 2).

vRouter Service diagram

This is game changing performance. Consider the software-only vRouter data path that supports roughly 1Mpps per x86 core. Customers tend to prefer deploying no more than four (4) x86 cores to the task of vRouter datapath in attempts to preserve as many cores as possible for virtual machines (VMs) and applications. This common, software-only configuration allows for 4Mpps of vRouter forwarding and traffic availability to VMs. When counting for bi-directionality, this translates to 2Mpps of receive traffic and 2Mpps of transmit traffic for VMs. The Agilio vRouter solution improves on the software-only performance appreciably, supplying up to 27Mpps of vRouter forwarding (when counting for bi-directionality, this translates to 13.5Mpps of receive traffic and 13.5Mpps of transmit traffic).

From a functionality perspective, the demonstration targets service chaining for VNFs as the key use case. The service chain will utilize MPLS over UDP to encapsulate L3 overlay network traffic sourced from a VNF and forward to a downstream VNF. Once received by the downstream system, the underlay network headers and MPLS over UDP encapsulation are removed (decapsulation) and forwarded to the host VNF over SR-IOV virtual functions (VFs). There are three key system benefits to this solution: 1) CPU cores are recovered through vRouter offload, 2) the amount of traffic that can be supplied to VNFs is drastically increased (from 4Mpps to 27Mpps with Agilio), and 3) VNFs can get native hardware I/O performance through the use of SR-IOV and zero-copy drivers with DPDK.

MPLS diagram

If you are going to be at MWC, we encourage you to stop by the Juniper booth (Hall 2, Stand 2J60) and Netronome’s hospitality suite for more details on the solution and benefits. If you can’t make it to MWC, please stay tuned for a future webinar that goes into more detail on the solution, product details, and use cases. Thanks and have a great week at MWC 2016!