An Investigation on Data Plane Programmability and Acceleration Methods
MetadataShow full item record
Programmability in networking may span multiple layers of the protocol stack. As cloud networking with resource management (orchestration of workloads), and automation evolved, hardware in computer networks also has also adapted to support better isolation principles while supporting forwarding functions. In this thesis, generic hardware such as servers was assumed to take on the role of legacy switching functions. In order to perform at the line rate, many data plane acceleration methods have been explored. We provided integration of one such data plane acceleration method, namely netmap, to the modular software switch initiative, xDPd (eXtensible Data Plane daemon). We present the measurement results with achieved improvements at levels of about 20% or more. We also present future directions in this paradigm shift.