Incentives - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 113

q i (d a ) q j (d c ) q i (d b ) q k (d c )

d a d c d b

t t + time

Downlink broadcast traffic (server) Query arrives; Send request (client)

FIGURE 5.18: System model for wireless data access [Yeung and Kwok,
2006a].

j and k request the same data item, d c , in the second interval. However,
the server is required to broadcast the content of d c only once in the next
broadcast period, which reduces the bandwidth requirement.
To successfully complete a query, a client expends its energy in two dif-
ferent parts: (1) informing the server of the desired data item, E U L ; and (2)
downloading the content of the data item from the common broadcast chan-
nel, E DL . The energy cost of sending a request to the server is represented by
E s . If a client sends a request for each query, then E U L would be the same
as E s . However, we show that clients need not send requests for each query
even without caching. This implies that E U L does not necessarily equal E s .
In general, the total energy required to complete a query, E Q , is given by:

E Q = E U L + E DL (5.43)

It is assumed that E DL is the same for all clients, but its value depends
on the size of a data item. In practice, E s is a function of various quantities,
including spatial separation, speed, instantaneous channel quality, bit-error
rate requirement, etc. For simplicity, however, E s is also assumed to be a
fixed quantity.
Based on this model of wireless data access, a novel utility function for
quantifying performance is defined as follows:

E total
E i Q

U i = (5.44)

where E total is the total energy available. The objective is then to reduce
the amount of energy consumed in the query process such that every client's
utility (Equation (5.44)) is increased.
Based on the utility function, Yeung and Kwok formulate the wireless data
access scheme as a non-cooperative game—wireless data access (WDA) game.
Game theoretic analysis shows that while the proposed scheme does not rely
on client caching, clients do not always send requests to the server. Simulation