Re: cross database joins

In article <4sgeps$5ra_at_athos.cc.bellcore.com> dmarcus_at_notes.cc.bellcore.com (Dave Marcus) writes:
>Path:
>nntp.teleport.com!news.serv.net!solaris.cc.vt.edu!newsfeed.internetmci.com!athos
>.cc.bellcore.com!news
>From: dmarcus_at_notes.cc.bellcore.com (Dave Marcus)
>Newsgroups: comp.databases.oracle
>Subject: Re: cross database joins
>Date: 16 Jul 1996 16:08:28 GMT
>Organization: Bellcore
>Lines: 23
>Message-ID: <4sgeps$5ra_at_athos.cc.bellcore.com>
>References: <4s6dl5$qvp_at_news-2.csn.net> <31E730CA.782E_at_quarks.idiscover.co.uk>
>NNTP-Posting-Host: oracle6-pc.cc.bellcore.com
>Mime-Version: 1.0
>Content-Type: Text/Plain; charset=US-ASCII
>X-Newsreader: WinVN 0.99.7

>In article <31E730CA.782E_at_quarks.idiscover.co.uk>,
>psx-info_at_quarks.idiscover.co.uk says...
>>
>>vance wrote:
>>>
>>> Is it possible to have a join drawing from tables that are in two different
>>> database on different machines in a TCP/IP network?
>>
 

>>...but be careful - I did something similar in Access to Oracle, mixing
>>local and remote tables - the query ran like a dog.
>>
>The reason they're dogs is because, oracle returns ALL of the data from the
>remote query into your local database first. Then it handles the where
>criteria. So if the remote table has for example 20000 rows, all 20,000 rows
>are returned into memory at the local site.

It depends on the sites too. I do joins between a 8 processor sparc and a 6 processor MVS machine and get very reasonable performance for sane table sizes (sane = <100k rows). you can also use tricks like referencing a view at the remote site that has limiting qualifications embedded in it. In that case the filtering occurs before the data is returned. Received on Tue Jul 16 1996 - 00:00:00 CEST