Re: Indexing techniques for hierarchies
Date: 21 Dec 2001 09:30:21 -0800
Message-ID: <a6e74506.0112210930.12830fe2_at_posting.google.com>
> ...different ways to represent hierarchy in the db
> ...how different methods fare from the performance perspective...
In believe oodbs are ideal at representing hierarchies and performing recursive algorithms. An oodb can be 10 to 100 times faster than rdbs in the following scenario:
In oo terms:
1. Create 1,000 classes (ie Cls1, Cls2, ...)
2. Each class has five properties:
(id int32, p2 float64, p3 bool, p4 text255, p5 dateTime)
3. Create an instance of each class.
4. Each object is the child of the previous.
Determine time to access/copy the value of each property of all objects to local variables starting at either the parent or the last child.
In relational terms:
(pk int32, p2 float64, p3 bool, p4 text255, p5 dateTime)
(Would need additional field fk start from T2 to relate records)
3. Create one record in each table.
1. Create 1,000 tables (ie T1, T2, ...)
2. Each record has five fields:
4. Each record is the child of the record in previous table.
Determine time to access/copy the value of each field of all records to local variables starting at either the parent or the last child.
I can provide code/benchmarks for an oodb. Would someone be willing to provide benchmarks for a rdb? Received on Fri Dec 21 2001 - 18:30:21 CET