ABSTRACT
In this chapter we introduce a new hybrid push-pull scheduling strategy. In short, the strategy partitions the entire set of items into two disjoint sets: the access set and the request set that are the push and pull subsets, respectively. The access set contains the hot data items that are the items that are very popular among clients and hence would be sent based on some push scheduling algorithm. The request set contains cold data items, which are not so popular items, and hence would be sent on-demand based on some pull scheduling algorithm as shown in Figure 3.1. The schedule strictly alternates between a push and a pull operation to transmit all the data items. While initially the system operates on unit-length, homogeneous data items, the work is extended to also include the heterogeneous, variable-length items. The selection criteria for a push item is based on packet fair scheduling and a pull item is selected on the basis of most requested first (MRF) (for homogeneous items) and stretch-optimal scheduling (for heterogeneous items). The scheme is further enhanced to incorporate the role of client priorities to resolve the tie. Suitable performance modelling is done to analyze the average system performance. Simulation experiments support this performance analysis and point out the efficiency of the hybrid system in reducing the overall average access time.
