Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Session: 1
Room: Park Tower 8219
Travis Machacek - New Mexico State University
Co-Author(s): Abderrahmen Mtibaa, Satyajayant Misra, NMSU ; Reza Tourani, St. Louis University
To date, there have been limited experimental comparisons of Named Data Networks (NDN) in different wireless configurations. For wider adoption of NDN for Internet of Things (IoT), this is an important subject of study. In this poster, we attempt to perform the first such experimental comparisons. Our experiments include NDN tunneled over IP and running directly over MAC. We anticipate that it is preferable to run NDN in ad-hoc mode over 802.11 due to the topology and the differences in transport protocols. Conceptually, NDN architecture is more favorable for implementing scalable IoT networks rather than the Internet Protocol (IP). Researchers have largely explored the 802.11 infrastructure mode deployment and often tunneling over an access point (AP). In this poster, we qualitatively and quantitatively compare NDN’s performance over both Layer 3 and directly on Layer 2 using infrastructure and ad-hoc modes of 802.11. We use a Raspberry Pi multi-hop wireless testbed which we configure using different wireless modes and network topologies. In this poster, we have deployed our nodes in a line topology and test data exchange between a source-destination pair (single flow) where NDN interest/data packets are sent directly over Layer 2 and Layer 3. We evaluate NDN?s performance over six and one hop networks. For each protocol, one run consists of sending one interest and measuring the time to receive data. We conduct 40,000 runs for each protocol/topology. To represent the latency characteristics for each protocol, we use the Cumulative Distribution Function (CDF). In ad-hoc mode, NDN over MAC is able to deliver 30% more packets under 75ms than NDN tunneled over IP. However, for larger latencies, which represent 2% of packets, NDN over MAC performs as good as UDP; NDN over TCP seems to have 4 to 5% more packets with 200ms or less compared to MAC and UDP. In infrastructure mode, we can see that NDN over ad-hoc mode is able to send up to 10% more packets with 55ms or less compared to NDN over infrastructure. Sending packets over infrastructure mode induces additional Layer 2 hops to and from access points which may add additional delays. Furthermore, MAC has a packet size < 1500B whereas UDP/TCP have a packet size > 1500.
We have shown that, in the ad-hoc scenario, running NDN over Layer 2 achieves better results reaching up to 30% latency reduction compared to Layer 3 deployment. In the future, we will perform further tests on multiple network topologies and realistic applications with multiple flows.
References: G. Grassi, D. Pesavento, G. Pau, R. Vuyyuru, R. Wakikawa and L. Zhang, ‘VANET via Named Data Networking,’?IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), Toronto, ON, 2014, pp. 410-415.
B. Han, X. Wang, N. Choi, T. Kwon and Y. Choi, ‘AMVS-NDN: Adaptive mobile video streaming and sharing in wireless named data networking,’ IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS), Turin, 2013, pp. 375-380.
Funder Acknowledgement(s): Funding was provided by the NSF CNS and the Intel Corporation.
Faculty Advisor: Satyajayant Misra, misra@cs.nmsu.edu
Role: In this research, I configured the Raspberry Pi testbed and wrote all of the code necessary to evaluate the performance of NDN over Layer 2 and Layer 3. I performed all of the tests in this work for each topology and analyzed the results from each test.