Querying Data under Access Limitations

Abstract

Abstract-Data sources on the web are often accessible through web interfaces that present them as relational tables, but require certain attributes to be mandatorily selected, e.g., via a web form. In a scenario where we integrate a set of such sources, and we pose queries over them, the values needed to access a source may have to be retrieved from other sources that are possibly not even mentioned in the query: answering queries at best can then be done only with a potentially recursive query plan that gets all obtainable answers to the query. Since data sources are typically distributed over a network, a major cost indicator for the execution of a query plan is the number of accesses to remote sources. In this paper we present an optimization technique for conjunctive queries that produces a query plan that: (1) minimizes the number of accesses according to a strong notion of minimality; (2) excludes all sources that are not relevant for the query. We introduce Toorjah, a prototype system that answers queries posed on sources with limitations by means of optimized query plans. Toorjah adopts a strategy that is aimed to retrieve answers as early as possible during query processing, and to present them to the user as they are computed. We provide experimental evidence of the effectiveness of our optimization, by showing the reduction of the number of accesses in a large number of cases. I. INTRODUCTIONIn the context of integration of web data [1], or in a data exchange setting where source data are retrieved on the web, information is often accessible only via forms; it is easy to see that accessing data through a web form amounts to querying a relational table, where a selection is specified by the fields that are filled in. Typically, certain fields are required to be filled in by the user in order to obtain a result; for example, all online shops forbid a request posed by a user who leaves all fields of the search form empty. Analogously, in legacy systems where data are scattered over several files and wrapped as relational tables, similar limitations are enforced.Limitations on how sources can be accessed significantly complicate query processing: as shown, e.g., in [2], [3], query answering in the presence of access limitations in general requires the evaluation of a recursive query plan. This is shown in the following example.Example 1: Suppose we have three relations: r 1 (Artist, Nation, YOB ), which stores artists with their nationality and year of birth, and requires the first attribute to be selected; r 2 (Title, Year , Artist), which stores data

Processing Keyword Queries Under Access Limitations

“…We represent Deep Web sources as relational tables; the necessity of filling in the aforementioned form with values forces one to access the relations with suitable selections; these restrictions are referred to as access limitations. It is well known that answering relational queries under access limitations requires the execution of a recursive query plan [5,4]. This is also true in our setting, where a recursive extraction of tuples (or facts) is performed, according to the access limitations, so as to search for the keywords in the data.…”

“…Therefore, query answering requires in general two main steps, described in Algorithm 1: (i) extract the reachable portion s of r; (ii) if possible, remove tuples from s so that the obtained set satisfies the conditions of Definition 1, while minimizing its size. A simple way of extracting the reachable portion, inspired by the procedure described in [4], is shown in Algorithm 2. This algorithm may be allowed to terminate early if the answer is not required to be optimal (flag ω set to f alse), and thus can stop as soon as the reachable portion contains all the keywords in the query.…”

“…The extraction of the reachable portion of an instance r with access limitations can be implemented by a Datalog program over r [4], which can be evaluated in polynomial time in the size of the input [7]. In addition, in [6] it is shown that the optimal q-fragments of r can be enumerated in ranked-order with polynomial delay, i.e., the time for printing the next optimal answer is again polynomial in the size of r. Hence, we can state the following preliminary result.…”

Cooperative Data Access in Multi-cloud Environments

“…Processing distributed queries under protection requirements has been studied in [9,12,14]. In these works, each relation/view is constrained by an access pattern, and their goals are to identify the classes of queries that a given set of access patterns can support.…”

Optimization of multi-domain queries on the web

“…For instance, some optimizations to be made during query plan generation to minimize the accesses to data sources are discussed in [11] for a subset of conjunctive queries, named connection queries; more expressive classes of queries, including conjunctive queries, are covered in [5]. Optimizing query plans.…”