A Distributed Dynamic Super Peer Selection Method Based on Evolutionary Game for Heterogeneous P2P Streaming Systems

Algorithm 1

A distributed 𝑄-learning. Super Peer Selection algorithm for ESS (ESS-SPS).

Step 1: Given the slot index 𝑡=0, each peer initializes 𝑥_𝑖(𝑡)=𝑥_𝑖(0) with a random between [0,1].

Step 2: Initialize 𝑄-value associated with 𝑄(𝑡)=0 for all peers in super groups.

Step 3: loop

      During slot 𝑡, for k = [1 : M]

      if  𝑥_𝑖(𝑡)≤𝛾  then

     peer 𝑖 randomly choose to be a SP to serve as a super peer to download streaming chunks

    from the peers outside the group with dowpnload rate 𝑟_𝑖(𝑡,𝑘) or to be an OP to download

    streaming chunks from the SP.

     else

    Choose strategy

     𝑏^∗=argmax_𝑏𝑄(𝑘)

     end if

    peer 𝑖 computes the indicator function 𝑜_𝑖(𝑡,𝑘) and his/her directly payoff  𝜋_𝑖(𝑡,𝑘) using (23).

Step 4: peer 𝑖 computes average payoff  𝜋_𝑖𝐵,𝑥_−𝑖(𝑡) and 𝜋_𝑖𝑥_𝑖(𝑡).

Step 5: Updates the probability of being a SP 𝑥_𝑖(𝑡+1) using (22).

Step 6: Update 𝑄(𝑘+1)=(1−𝜆)𝑄(𝑘)+𝜆𝜋_𝑖(𝑥_𝑖(𝑡))+𝛽max_𝑏𝑄(𝑘)

Step 7: endloop for all peers in super group.