Flea Net A Virtual Market Place on

Flea. Net : A Virtual Market Place on Vehicular Networks Uichin Lee, Joon-Sang Park Eyal Amir, Mario Gerla Network Research Lab, Computer Science Dept. , UCLA

Advent of VANETs n Emerging VANET applications ¨ Safety driving (e. g. , Traffic. View) ¨ Content distribution (e. g. , Car. Torrent/Ad. Torrent) ¨ Vehicular sensors (e. g. , Mob. Eyes) n What about commerce “on wheels”?

Flea Market on VANETs n Examples ¨A mobile user wants to sell “i. Pod Mini, 4 G” ¨ A road side store wants to advertise a special offer n How to form a “virtual” market place using wireless communications among mobile users as well as pedestrians (including roadside stores)?

Outline Flea. Net architecture n Flea. Net protocol design n Feasibility analysis n Simulation n Conclusions n

Flea. Net Architecture -- System Components n n Vehicle-to-vehicle communications Vehicle-to-infrastructure (ad-station) communications * Roadside stores (e. g. , a gas station)

Flea. Net Architecture -- Query Formats and Management n Users express their interests using formatted queries ¨ e. Bay-like n n category is provided E. g. , Consumer Electronics/Mp 3 Player/Apple i. Pod Query management ¨ Query DB) storage using a light weight DB (e. g. , Berkeley ¨ Spatial/temporal queries ¨ Process an incoming query to find matched queries (i. e. , exact or approximate match) n E. g. Query(buy an i. Pod) Query(sell an i. Pod)

Flea. Net Protocol Design n Flea. Net building blocks ¨Query dissemination ¨Distributed query processing ¨Transaction notification n Seller and buyer are notified n This requires routing in the VANET challenges ¨Large scale, dense, and highly mobile n Goal: designing “efficient, scalable, and non-interfering protocols” for VANETs n

Query Dissemination Yellow Car w/ Q 1 n n Q 2 Query dissemination exploiting vehicle mobility Query “originator” periodically advertises its Red Car w/ Q 2 query to 1 -hop neighbors Q 1 Q 2 ¨ Vehicles “carry” received queries w/o further relaying Q 1

Distributed Query Processing Q 2 QM QM Cyan car w/ QM n (1) Find a matching query for Q 2 No match found (2) Send a match notification msg to the originator of query QM Received query is processed to find a match of interests (1) Find a matching Local. Match Q 2 QM Q 1 ¨ Eg. query for Q 1 – buy i. Pod Q/ Q M – sell i. Pod / Q 2 – buy Car Q Found query M 1 QM Red car w/ Q 2 & carries Q 1

Transaction Notification n After seeing a match, use Last Encounter Routing (LER) to notify seller/buyer ¨ Forward a packet to the node with more “recent” encounter QM Local. Match QM Q 1 Cyan car Q 1 T-1 s Green car Yellow car TRX RESP TRXREQ Blue car Q 1 T-5 s Red car Q 1 T-10 s Encounter timestamp Q 1 T Originator of Q 1 Current Time: T

Flea. Net Latency n n Restricted mobility patterns are harmful to opportunistic data dissemination However, latency can be greatly improved by the popularity of queries Popularity distribution of 16, 862 posting (make+model) in the vehicle ad section of Craigslist (Mar. 2006) Frequency (log) n Items (log)

Flea. Net Scalability n n Assume that only the query originator can “periodically” advertise a query to its neighbors We are interested in link load Load depends only on average number of neighbors and advertisement period (not on network size) Example: ¨ Parameter setting : R=250 m, 1500 B packet size, BW=11 Mbps ¨ N=1, 000 nodes in 2, 400 m x 2, 400 m (i. e. , 90 nodes within one’s communication range) ¨ Advertisement period: 2 seconds ¨ Worst case link utilization: < 4%

Simulations n n Ns-2 network simulator 802. 11 b - 2 Mbps, 250 M radio range Two-ray ground reflection model “Track” mobility model ¨ Vehicles move in the 2400 mx 2400 m Westwood area in the vicinity of the UCLA campus n Metric ¨ Average latency: time to find a matched query of interest Westwood area, 2400 mx 2400 m

Simulation Results n Impact of density and speed

Simulation Results n Impact of query popularity Popularity: the fraction of users with the same interest ¨ For a single buyer, increase the number of sellers (e. g. , N=200/0. 1 = 20 sellers) ¨

Simulation Results n Impact of ad-station location Given N=100, fix each node in its initial location, and set it as a “stationary” ad-station (as a buyer) ¨ measure the average latency to the remaining 99 mobile nodes (run 99 times, by taking turns as a seller: 1 buyer 1 seller) ¨ N=100/V=25 m/s avg. stationary avg. mobile Latency rank

Conclusions n Proposed a virtual market concept in VANETs: ¨ n A mix of mobile and stationary users carry out buy/sell transactions (or any other matching of common interests) using vehicular networks Mobility-assisted query dissemination and resolution (scalable and non-interfering) Node density/speed are closely related to the performance ¨ Popularity of a query greatly improves the performance ¨ Location of an ad-station is important to the performance ¨ n Future work ¨ Query aggregation to improve the performance n Unpopular queries/queries from ad-stations How to enforce cooperativeness of users? ¨ Security: false query injection and spamming? ¨