Loading...

Packet Processing Acceleration for Virtualized Network Functions

Barati Sedeh, Amir Reza | 2021

159 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54704 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Jahangir, Amir Hossein
  7. Abstract:
  8. Network Function Virtualization (NFV) is a new paradigm in computer networks for implementing network functions on general-purpose common-of-the-shelf processors. The revolutionized NFV architecture has shown to be a suitable replacement for hardware middleboxes. NFV is in high demand because of its lower costs, increased flexibility, and more scalability. However, the ever-growing volume of network traffic and increased data transfer speed in computer networks have made achieving high performance the major issue in this area. With thousands of processing cores and high memory bandwidth, Graphical Processing Units (GPUs) are potentially an appropriate solution to accelerate network functions. GPUs can provide fine-grained parallelism for processing network packets according to their unique architecture. However, using this accelerator comes with two major challenges: the delay overhead of transferring packets and the throughput degradation due to co-execution and change in the traffic pattern. In this thesis, we propose an adaptive load balancer to mitigate the delay overhead of transferring packets and a software scheduler based on persistent threads with dynamic resource partitioning capability to reduce the negative effects of co-execution and change in the traffic pattern. Our experiments show that our adaptive load balancer achieves a 29% and 145% increase in throughput compared to the GPU-only and CPU-only designs, respectively. Also, our software scheduler with dynamic SM partitioning capability increases the throughput by 13% relative to static partitioning and by 20% compared to the Hyper-Q method
  9. Keywords:
  10. Graphics Procssing Unit (GPU) ; Virtualized Network Functions ; Adaptive Load Balancing ; Software Scheduler ; Dynamic Partitioning

 Digital Object List

 Bookmark

No TOC