M. Stonebraker. The case for partial indexes. In SIGMOD Record, volume 18, 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....were not useful for any of the queries. Having large indices adversely impacts the performance of the vertical representation because of the time spent in loading the indices and the increase in path length due to deeper index trees. It would have helped if the database supported partial indices [20] that allow only the rows of interest to be indexed. 4.4 Performance Results We now summarize the important results from a very large number of experiments we performed. We will present the results for project, select, join and aggregation operations. We flushed buffer pool, main memory and file ....

....Horizontal Vertical Binary Manageability Flexibility Performance ffl Partial indices We create an index on each of the three columns of the vertical table. In the process, we end up indexing the entire data in the vertical table. Having database support for partial indices [20] that allow only the rows of interest to be indexed will help improve the performance of the vertical representation. ffl Enhanced table functions The table function syntax needs to be extended with additional clauses to specify the required order of input arguments. This facility is critical for ....

M. Stonebraker. The case for partial indexes. SIGMOD Record, 18(4):4--11, 1989.

....(this is a hybrid of P S and Index Stride) but this will involve several random I Os. Alternatively, we can store tuples from these rare groups in a separate table, and have a clustered index on the group by column on this table this is similar to a partial index [Seshadri and Swami 1995; Stonebraker 1989], except that these tuples are now clustered separately and therefore one avoids multiple random I Os. The challenge lies in automatically deciding which values of which column(s) to treat in this manner, taking into account the frequency of queries that will use this table, as well as the ....

Stonebraker, M. The case for partial indexes. SIGMOD Record, 18(4), 1989.

....most preferred groups alone (this is a hybrid of P S and Index Stride) but this will involve many random I Os. Alternatively, we can store tuples from these rare groups in a separate table, and have a clustered index on the group by column on this table this is similar to a partial index [Sto89] except that these tuples are now clustered separately and therefore one avoids multiple random I Os. The challenge lies in automatically deciding which values of which column(s) to treat in this manner, taking into ac 0 500 1000 1500 2000 2500 0 10 20 30 40 50 60 70 80 90 100 Number of ....

M. Stonebraker. The case for partial indexes. In SIGMOD Record, 18(4):4--11, 1989.

....O O O O O O . O O O O O O . COLUMNS TUPLES Partial Indexes on C 1, C 2, C 3, C 4 and C 5 Indexed regions Figure 2: Partial Indexes COLUMN DOMAIN VALUES FREQUENCY OF OCCURRENCE Indexed regions Figure 3: Using Distributions In [9], a case is made for adding partial indexes to a database system. It assumes that the nature of the query predicates is known. For instance, if it is known that queries will only contain predicates of the form X 200, it is sufficient to build an index on column X only for values greater than ....

....adding partial indexes to a database system. It assumes that the nature of the query predicates is known. For instance, if it is known that queries will only contain predicates of the form X 200, it is sufficient to build an index on column X only for values greater than 200. It is suggested in [9] that this constraint be specified by the DBA as a clause in the command that creates the index. Since the partial indexing of [9] which has been implemented in POSTGRES [7] is one of the several possible partial indexing strategies described in this paper, we call our strategy generalized ....

[Article contains additional citation context not shown here]

M. Stonebraker. The Case for Partial Indexes. ACMSIGMOD Record, 18(4):4--11, December 1989.

.... continue queries (or pieces of queries) that they have previously canceled. Techniques for checkpointing the state of online aggregation queries should therefore prove very useful. Checkpoints of partially computed queries can also be used as materialized sample views [Olk93] or partial indexes [Sto89], depending on their makeup. Query 4: SELECT SUM(s.sale price) v.total sales FROM sales s, SELECT SUM (sales.sale price) AS total sales FROM sales) v, GROUP BY s.dept; 14 5.5 Concurrency Control Online aggregation queries are likely to be very time consuming. Such queries will typically ....

Michael Stonebraker. The Case for Partial Indexes. SIGMOD Record, 18(4):4-11, 1989.

....disk arm or collection of arms. These indexes would have the form: index on EMP(salary) where age 20 index on EMP(salary) where age 40 index on EMP(salary) where age = 20 and age = 40 Such data structures are partial indexes and offer a collection of desirable features as discussed in [STON88]. Notice that such indexes are smaller than a current conventional index and should be contrasted with other proposals (e.g join indexes [VALD87] links [ASTR76] and the indexes in IMS [DATE84] which are larger than a conventional index. Three issues must be addressed by a parallel optimizer in ....

Stonebraker, M., "The Case for Partial Indexes," Electronics Research Laboratory, University of California, Berkeley, CA, Report ERL M88/62, June. 1988.

No context found.

M. Stonebraker. The case for partial indexes. In SIGMOD Record, volume 18, 1989.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC