ABSTRACT
In most practical systems, clients often become impatient while waiting for the designated data item. After a tolerance limit, the client may depart from the system, resulting in a drop of access requests. This behavior significantly affects the system performance, which needs to be properly addressed. There are also ambiguous cases that reflect the false situation of the system. Consider the scenario in which a client gets impatient and sends multiple requests for a single data item to the server, as shown in Figure 5.1. Even if that particular data item is not requested by any other client, its access probability becomes higher. In existing systems, the server remains ignorant of this fact and thus considers the item as popular and inserts it into the push set or pull it at the expense of some other popular item. In contrast, our work [43] reduces the overall waiting time of the system taking care of such anomalies.
