1 / 19

Seung-Hoon Lee, Uichin Lee, Kang-won Lee*, Mario Gerla University of California, Los Angeles

Content Distribution in VANETs using Network Coding: The Effect of Disk I/O and Processing O/H. Seung-Hoon Lee, Uichin Lee, Kang-won Lee*, Mario Gerla University of California, Los Angeles IBM T. J. Watson Research Center*. Content Distribution in Vehicular Ad-Hoc Networks (VANETs).

Download Presentation

Seung-Hoon Lee, Uichin Lee, Kang-won Lee*, Mario Gerla University of California, Los Angeles

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Content Distribution in VANETs using Network Coding: The Effect of Disk I/O and Processing O/H Seung-Hoon Lee, Uichin Lee, Kang-won Lee*, Mario Gerla University of California, Los Angeles IBM T. J. Watson Research Center*

  2. Content Distribution in Vehicular Ad-Hoc Networks (VANETs) • Applications • Software updates and patches (e.g., navigation map, games) • Multimedia data downloads (e.g., videos, news, etc.)

  3. Content Distribution Challenges • High mobility (i.e., highly dynamic networks) • Error-prone channel (due to obstacles, multi-path fading, etc.)

  4. CarTorrent: BitTorrent-like Cooperative Content Distribution in VANETs A file is divided into pieces Web Server Exchange pieces via Vehicle-to-Vehicle Communications Download a file (piece by piece) Not useful! • Problem: Peer & Piece selection  coupon collection problem Cannot complete download!

  5. Using Network Coding: CodeTorrent 1 more? A file is divided into pieces Web Server Must consider network coding O/H!! Any linearly independent coded packet is helpful • Network coding “effectively” mitigates the peer/piece selection problem and “improves” the performance!

  6. Worst Case Scenario Coded piece is not delivered! Request Network coding must be carefully configured! Should investigate “origin” of network coding O/H!!

  7. Closer Look at Network Coding O/H Request

  8. Closer Look at Network Coding O/H Check memory Cache Missing (Memory is limited) Disk I/O X1 X2 X3 Blocks loaded to Memory e1 e2 e3 CPU O/H (Network Coding) + Coded Block generation [e1,e2,e3] e1X1+e2X2+e3X3

  9. Network Coding O/H Model • Overall process: Reading (cache miss)  Encoding  Sending • Parallelization is feasible: Send an encoded “packet” ASAP • Disk access O/H: • Must read all the necessary data before encoding • Disk input rate is determined by the characteristics of a disk • O/H is proportional to the number of pieces to be encoded • Encoding O/H • Per symbol encoding (∑eiXi) O/H: Linearly proportional to the cost of a pair of Galois Field (GF) operations (multiplication & addition) • Encoding rate = (# pieces * GF <+,*> time)-1 • O/H is proportional to the number of pieces to be encoded

  10. 1 4 Mitigating Coding Overheads • Solutiion#1: divide a file into small generations • Problem: too many generation causes a coupon collection problem • Conflicting goals: maximizing benefits of NC vs. minimizing coding O/H 10MB x 5 50MB

  11. Gen 3 Request Gen 3 Ex. Sparse Coding Rate: 50% Need Gen 1,3, 5,7 Mitigating Coding Overheads • Sol#2: encode only a fraction of pieces (i.e., sparse coding) • Problem: how to find the right value? • Sol#3: pull a generation that is in the buffer of a remote node (remote buffer aware pulling)

  12. Evaluating Impacts of Coding O/H • Difficult to evaluate the overall impacts of coding O/H in VANETs • Dynamic nature of VANETs (high mobility) • Large scale scenarios • Network Simulator (e.g., NS2, QualNet) only models communication O/H • Implement our measurement based models into a network simulator (QualNet)

  13. Simulation Setup • Communications • 802.11b; 11Mbps + Two-ray ground propagation • Mobility • Real-track model w/ speed range of [0,20] m/s • UCLA area map: 2400m*2400m • Nodes • 3 APs: file sources • 200 nodes/40% interest level: • 80 nodes are downloading a file • O/H model • Coding rate: 7.6 Mbps • Disk I/O rate: 38 MB/sec

  14. Simulation Results (1) • Delay without O/H • Small # of generations is a better choice • Larger # of generations  more severe coupon collection problem Gen1 Gen5 Delay without O/H

  15. Simulation Results (2) • Delay with O/H (Buffer 50%: CPU O/H + Disk I/O) • Small # of generations is not a better choice any longer!! • Single generation scenario is even worse than “No coding” case. • Must carefully choose the number of generations! Delay with O/H (Buffer 50%)

  16. Simulation Results (2) “U” shape The optimal point exists Delay with O/H (Buffer 50%) • Delay with O/H (Buffer 50%: CPU O/H + Disk I/O) • Small # of generations is not a better choice any longer!! • Single generation scenario is even worse than “No coding” case. • Must carefully choose the number of generations!

  17. Simulation Results (3) • Sparse coding rate must be carefully chosen. Delay with Sparse Coding (50MB)

  18. Simulation Results (4) • Remote Buffer Aware Pulling (RBAP) • Successfully reduces disk I/O O/H Delay with RBAP (50MB)

  19. Conclusion • Investigated the impacts of network coding O/H • Disk I/O + Processing O/H • Designed “measurement” based models • Evaluated various strategies to mitigate O/H • Multiple generations, sparse coding, buffer aware pulling • Lessons learned: network coding must be carefully configured to maximize its benefits • Future work • Good configuration? – must tune various factors, i.e., piece size, disk access/coding rate, and shared bandwidth • Understand the impacts of O/H and study enhancement techniques (e.g., H/W acceleration) in various environments (e.g., embedded systems, Smart Phones)

More Related