Datapath Acceleration
The paradigm shift from “bump in the wire” network functions towards virtualized network functions, and an increasingly high density of data traffic, necessitates acceleration in the data path of the general purpose CPUs, and new ways to minimize packet processing overhead to match the hardware capacity.
With the advent of NFV and big data, the need for fast packet processing is at an all-time high. NFV, for its success in business, relies on the data-path performance of virtual machines hosting the functions.
To facilitate high performance in the data path, a software framework known as Intel DPDK (Data Plane Development Kit) has been provided by Intel to help users create a sustainable, high-density traffic across network function on x86 platforms.
xFlow Research Inc. has developed an API to abstract and simplify application development on Intel DPDK. This API is named as Twister. You can download Twister API for free from its website at the following link:
https://xflowresearch.com/twister
xFlow Research Inc. has been providing support and services for datapath accelerations in the following domains:
- We have developed proofs of concept of high velocity clouds for our customers in which we successfully accomplished the provisioning of network functions having typical traffic throughput with 10x the performance of traditional Linux IP stack.
- We have expertise in kernel-bypass networking that allowed the control of ethernet hardware to user space – bypassing the overhead of OS kernel.
- We have successfully accomplished the deployment of dedicated network functions, with fast underlying datapath, in load-intensive platforms.