Skip navigation

Improving the Flexibility and Performance of Virtual Network Functions

Event Date: September 28, 2017
Speaker: Dr. Bo Han
Speaker Affiliation: Principal Inventive Scientist, AT&T Labs - Research
Sponsor: Purdue School of Industrial Engineering
Time: 10:00am
Location: GRIS 316
Priority: No
College Calendar: Show


Network Functions  Virtualization (NFV)  was  recently  proposed  to  improve  the efficiency of  network  service  provisioning,  by leveraging virtualization technologies  and  commercial off-the-shelf  hardware.  I t  transforms  how network operators  architect  their  infrastructure by decoupling the software implementation of network functions from the underlying hardware . However, the current commercial implementations of Virtual  Network Functions (VNF) are not cloud­ friendly.  There  is  still  a  gap  between  their  performance  and  that  of the  physical network functions.  This talk will start with a brief overview  of  NFV,  introducing its design principles and research challenges.  It will then present two recent projects  at  AT&T,  EdgePlex  and  Para Box, on  improving the  flexibility and performance of  VNFs.  EdgePlex decomposes and virtualizes the functions of a traditional edge router and enhances its flexibility.  ParaBox is a novel hybrid packet processing architecture that reduces the latency of VNFs when they are provisioned in service chains.


Dr. Bo  Han is  a  Principal  Inventive Scientist at  AT&T  Labs - Research.   His current research interests are in the areas of network functions virtualization, software defined networking,  mobile computing,  and wireless networking, with a focus on developing simple but efficient and elegant solutions for real-world networking and systems problems.  He obtained his Bachelor's degree from Tsinghua  University,  his M.Phil.  from  City  University  of  Hong  Kong,  and  his Ph.D. from the University of  Maryland,  all in Computer Science.  He has published 50+ papers in  top-tier journals  and conferences.   He received the best paper award from ACM  CoNEXT  2016.