As shown in Figure 1.11, in a hybrid approach, we divide the set of data items in the server, into two disjoint sets, i.e., access set and request set. The access set contains the hot data items that would be sent based on some push scheduling algorithm. The request set contains cold data items that would be sent on demand based on some pull scheduling algorithm. Hot data items are the items that are very popular among clients; so it is assumed that they are requested by some clients (among the wide population of the clients) at all times. Cold data items are the data items that are not so popular among the clients. The cutoff point, K as shown in Figure 1.11, separating the items from the push set and the pull set, is chosen in such a way that the overall expected access time is minimized.
A detailed overview of the published research works on wireless data broadcast can be found in . Therefore, the search for efficient hybrid scheduling, which
FIGURE 1.11 Essence of hybrid scheduling.
explores the efficiency of both push and pull strategies, continues. Examples of hybrid push-pull systems include the Hughes Network System DirecPC Architecture , that uses satellite technology as shown in Figure 1.12 to give a fast, always-on Internet connection; the Cell Broadcast Service (CBS) that enables the delivery of short messages to all the users in a given cell in both GSM and UMTS systems ; and the Service Discovery Service in networks of pervasive devices .
The general notion of hybrid scheduling lies in dividing the entire set of data items into two parts: popular items and less popular items. The scheduler pushes the popular data items at regular intervals. It also accumulates the client's requests for less popular items in the pull queue and selects an item depending on the specific selection criteria. A wise selection of the cutoff point, used to segregate the push and pull sets, has the power to reduce the overall expected waiting time of the hybrid system. However, most hybrid scheduling is based on homogeneous (often unit-length) data items. The effect of heterogeneity, with items having different lengths, needs to be considered to get an efficient, hybrid scheduling strategy for asymmetric environments.
Was this article helpful?