WIP: master-slave seqscan patch - Mailing list pgsql-hackers

From Qingqing Zhou
Subject WIP: master-slave seqscan patch
Date
Msg-id Pine.LNX.4.58.0604092235530.10639@eon.cs
Whole thread Raw
List pgsql-hackers
Upon several requests of the master-slave patch source code, I sent it
here. This is a proof-of-concept implementation only to show it is
feasible only.

To play with it, applied it to cvs tip. When you key in a SQL like:
SELECT max(i) FROM t; // t with more than 200000 records

You are suppoesd to key in the pid (printed on the screen) of jet engine
to make it running.

Now Jonah and I are working on two lines to make it better. Jonah will
work on the basic comunication part (extend this two more than one slaves)
and I am working on Para structure -- you can see the ExecInitBuffering()
is almost the same as the SeqScan, which is quite ugly. I will try to
cleanup this part and add support to IndexScan to show that it is
extensible.

Many bugs and problems are in this bloody implemantion. The xid problem,
error communication problems, etc are not solved. If you can see any
show-stop reasons of not doing so, please say it.

Regards,
Qingqing

pgsql-hackers by date:

Previous
From: Neil Conway
Date:
Subject: FYI: summer plans
Next
From: Peter Eisentraut
Date:
Subject: Updating OID column