Balancing Push and Pull

The work of Acharya, Franklin, and Zdonik [2] is perhaps the first significant work that effectively explores the advantages of both push- and pull-based data transmission strategies. The work introduces the asymmetry in different factors, like, (1) uplink and downlink channels, (2) clients and server ratio, and (3) amount of data downloaded and uploaded. The proposed strategy considers a capacity-constrained server and multiple clients with uplink channels. It then extends the static, push-based data broadcasting to incorporate pull-based on-demand data transmission schemes for read-only data items. For push-based broadcast scheduling, the proposed algorithm selects the cached page which contains lowest p/x ratio. The pull-based on-demand scheduling is modelled as a point-to-point connection with the server. While the rate of client-requests increases as the number of clients increases, the server only has a maximum number of allowable requests it can handle. The server is capable of interleaving push- and pull-based data items, and options are kept to vary the percentage of slots dedicated for on-demand pull scheduling. The requests are accumulated and kept in the pull queue. The server selects the item in a first-come-first-serve (FIFO) fashion. A threshold parameter is kept to maintain the use of the back channel under certain limits. While measured client models a single client, the virtual client models the combined effect of all other clients in the system. It maintains a cache holding different pages and waits for certain time units between two consecutive requests. If possible, the requests are serviced from the cache; otherwise they are broadcast or pulled.

Was this article helpful?

0 0

Post a comment