Index Stride provides a user with siumultaneously refining estimates for each group in a GROUP BY query. Without the access method, the distinct confidence intervals could not be refined concurrently, and the relative rates for the groups could not be controlled on the fly.

To understand the behavior of Index Stride, consider the following query:

SELECT ONLINE college, avg(grade), count(*) from grades

GROUP by college

Index Stride first probes the index to find all the groups (colleges). It then processes tuples from each group in a "round robin" schedule, improving each college's confidence interval and updating its running estimate. The user can manipulate the schedule at which different groups are processed by increasing, decreasing, or stopping the amount of time spend sampling each group as results arrive. The picture below shows an interface for a query that includes an Index Stride. The user can stop a group, increase processing time for another and view the updates for all groups simultaneously.


[ Home | Projects | News | Papers | People | Demos]


Send mail to with questions or comments about this web site.
Last modified: August 04, 1997