vinay@ai.toronto.edu ("Vinay K. Chaudhri") (10/19/90)
The idea of using predicate locks for database concurrency control has been around for over ten years. But still there are not many commercial systems which use predicate locking as a primary method for concurrency control. One reason for this lack of applicability is because it is not computationally inefficient to check overlap between two predicate locks. People have given suggestions for developing some heuristics but there does not seem to be any comprehensive approach available. Another difficulty seems to be in devising an efficient implementation technique for predicate lock table. Maintaining a predicate lock table of some reasonable size (say 100) and comparing a new predicate against these 100 expressions is prohibitive, leave aside the computational reasons. My question is the following: Are there really no commercial systems that use predicate locking ? Are the reasons for their in-applicability as cited above are correct ? Are predicate locks dead as a research or application area ? --------------------------------------------------------------------------- Vinay K Chaudhri Email: Mail: Internet: vinay@ai.toronto.edu 321 Bloor Street (West) Box 204 Bitnet : VINAY@UTORANT.BITNET Toronto M5S 1S5 UUCP : ...!uunet!utai!vinay CANADA
dmg@Unify.Com (Dave Glende) (10/20/90)
In article <90Oct18.180225edt.12896@neat.cs.toronto.edu> vinay@ai.toronto.edu ("Vinay K. Chaudhri") writes: > > >The idea of using predicate locks for database concurrency control >has been around for over ten years. But still there are not many >commercial systems which use predicate locking as a primary method >for concurrency control. > >One reason for this lack of applicability is because it is not >computationally inefficient to check overlap between two predicate >locks. People have given suggestions for developing some heuristics >but there does not seem to be any comprehensive approach available. > >Another difficulty seems to be in devising an efficient implementation >technique for predicate lock table. Maintaining a predicate lock >table of some reasonable size (say 100) and comparing a new predicate >against these 100 expressions is prohibitive, leave aside the >computational reasons. Using predicate locks as the sole concurrency control mechanism is, as you've said, quite expensive. It would be possible to use them in small applications, but in a typical commercial situation the cost becomes too high especially in worse-case scenarios. However, certain types of predicate locks can be quite useful and still cost effective. > >My question is the following: Are there really no commercial systems >that use predicate locking ? Are the reasons for their in-applicability >as cited above are correct ? Are predicate locks dead as a research >or application area ? > UNIFY 2000 uses a combination of predicate locks and row locking for concurrency control. -- David Glende Work:(916) 920-9092 | Unify Corporation dmg@unify.com FAX :(916) 921-5340 | 3870 Rosin Court | Sacto, CA 95834
gupyo01@cai.com (11/03/90)
In article <90Oct18.180225edt.12896@neat.cs.toronto.edu>, vinay@ai.toronto.edu ("Vinay K. Chaudhri") writes: > The idea of using predicate locks for database concurrency control > has been around for over ten years. But still there are not many > commercial systems which use predicate locking as a primary method > for concurrency control. > > [... stuff deleted ...] > > My question is the following: Are there really no commercial systems > that use predicate locking ? Are the reasons for their in-applicability > as cited above are correct ? Are predicate locks dead as a research > or application area ? > At least one commercial product uses predicate locks: CA-DB from Computer Associates. One of the reasons for this might be that the product was originally developed by a start-up which had Kapali Eswaran as one of its founders. For more information on this, please contact me. Yogesh Gupta Computer Associates (408) 922-2633