The social network sites has been very popular these days. In these systems, re sources are associated with each user (i.e., network node) in the network, and these resources include both soft and physical resources, such as pictures, videos, books, etc. Users would like to control who can see their resources, e.g., any one, only friends, friends and families, friends of friends, etc. The rules defining who can see the resources are called visibility views, and in our system they are identified

more »

... by the owner nodes and regular path expressions. The users can issue queries for discovering resources, e.g., "Find all lawnmowers in my three degree network that are visible to me". The evaluation process needs to check all candidate resources to see if they are visible to the query issuer, and this requires computing the visibility views associated with every candidate resources. The visibility views are expensive to evaluate. In order to facilitate the visibility query answering, it is necessary to pre-compute and materialize the views in the so called "cache". But because of the tremendous volume of the whole view data, we have to select only a portion of them to materialize, based on metrics such as the views' sizes, the costs to evaluate them, the popularity of references, the temporal locality of requested views, and the correlation relationships among them. The problem of trying to select the views in a static way is called the static view selection problem, and it has been studied extensively by the database community, mostly for answering OLAP queries in data warehouses. We call it "static" because it pre-computes and materializes the views in one lump-sum, based on statistics accumulated before the query sequence starts. This problem has a deep root from the caching problem studied in the theory and system communities. The researches done on caching policies solve the problem in an ad-hoc dynamic way by making decisions at each request, which make use of the temporal locality. Many ad-hoc policies have been p [...]