Current industry insight holds that programmable processors (both CPUs and NPUs) are of lower performance than their hard-coded counterparts, such as Ethernet chips.
In this paper we show that Open Flow (OF) can be used for implementing various network functions, such as routing, and the performance we can get can match the non-OF implementations of the given network function once some bottlenecks are removed. Our prototype can do 100 Gbps (gigabit per second) even with small packets on one Intel blade, reaching a noticeable 110 Mpps packet speed with 64 byte long packets.
Not yet available in IEEEXplore
Invited paper presentation 2015-07-01 at the 16th IEEE International Conference on High Performance Switching and Routing (HPSR 2015)
© 2015 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.