| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1065.1 | | NOVA::MOY | Michael G. Moy, Rdb/VMS Engineering | Tue Jan 21 1992 08:42 | 9 |
| Rdb supports horizontal partitioning of tables and indices within a
database if this is what you mean by horizontal partitioning. If, by
horizontal partitioning, you mean multi-database joins, then Rdb
doesn't have this.
Ingres has this (don't know about the others) in their IngresStar
product.
michael
|
| 1065.2 | Multiple tables in networked databases.. | YOSMTE::OLDING_NI | dtn 549 5928 | Tue Jan 21 1992 21:21 | 8 |
| Re .-1 Well, yes, multi-database joins. The ability to view data from
tables that exist in multiple physical databases across a network as
though they were in one table, with the associated optimizations needed
for performance (where the joins occur, where data is moved from and
to, etc).
Is that what Ingres claims to do? Anyone know if Sybase and Oracle are
claiming that they do?
N
|
| 1065.3 | Oracle YES, Sybase ? | KETJE::GERARD | Jean-Paul GERARD - EIS - Brussels | Wed Jan 22 1992 14:50 | 11 |
| Oracle does support horizontal partitioning thru the use of tablespaces mapped
to multiple datafiles on different spindles, however record clustering as you
could do it with Rdb is not supported, i.e. you do not control record occurence
location.
Sybase uses segments to map tables to datafiles, which is similar to an Rdb
multi-file database. I'm not sure you can link a table with one or more
segments. Should this be possible, then horizontal partitioning would be suppor-
ted by Sybase.
Jean-Paul
|
| 1065.4 | some thoughts | COPCLU::BRUNSGAARD | Curriculum Vitae, who's that ?? | Wed Jan 22 1992 15:52 | 21 |
| Also be were aware about the sematic of multi database since .-1 only
talks about multi-file principles within one database.
Ingres: Through Ingres Star a fair amount of distributed of queries
can be done with some optimisations, but not too much.
There are no rewrite of queries to better optimise for network acces,
just the plain old how-to-join-two-tables-from-what-you-wrote stuff
added with a little bit of network information.
Note that it is the bast there is released currently, so I am not sayig
it isnt good, just that it is sufficient for real production usage.
Also a pretty sofisticated handling of metadata changes within Star is
supported. However I don't really know how base table changes are
propageted into the star product (if at all aplicable ?)
Oracle: Since they don't have a single database optimiser how could
they have a distributed optimiser !?!
Their implementation is purely to be able to say. YES we have a *Star
product, but if you look into it is very close to useless
Sybase: I really don't know.
Lars
|
| 1065.5 | Rdbstar is real stuff | COPCLU::BRUNSGAARD | Curriculum Vitae, who's that ?? | Wed Jan 22 1992 15:56 | 13 |
| also have you heard about Rdbstar ??
This product has bee architected to support exactly that kind of
environment (ie multidatabase multivendor multinetworked).
Look in the notes file at
BROKE::RDBSTAR_DEC (press KP7 to add)
This is the real stuff !!!
Even though Ingres is better than Oracle, the product was not
architected that way, and that is bound to mean problem at some point
when using it.
Lars
|