Re: Master/Slave, DB separation or just spend $$$? - Mailing list pgsql-performance

From Chris Browne
Subject Re: Master/Slave, DB separation or just spend $$$?
Date
Msg-id 877hy0r8gf.fsf@dba2.int.libertyrms.com
Whole thread Raw
In response to Master/Slave, DB separation or just spend $$$?  (Kelvin Quee <kelvinq@gmail.com>)
List pgsql-performance
kelvinq@gmail.com (Kelvin Quee) writes:
> I will go look at Slony now.

It's worth looking at, but it is not always to be assumed that
replication will necessarily improve scalability of applications; it's
not a "magic wand" to wave such that "presto, it's all faster!"

Replication is helpful from a performance standpoint if there is a lot
of query load where it is permissible to look at *somewhat* out of
date information.

For instance, replication can be quite helpful for pushing load off
for processing accounting data where you tend to be doing analysis on
data from {yesterday, last week, last month, last year}, and where the
data tends to be inherently temporal (e.g. - you're looking at
transactions with dates on them).

On the other hand, any process that anticipates *writing* to the
master database will be more or less risky to try to shift over to a
possibly-somewhat-behind 'slave' system, as will be anything that
needs to be consistent with the "master state."
--
(reverse (concatenate 'string "ofni.secnanifxunil" "@" "enworbbc"))
http://www3.sympatico.ca/cbbrowne/spiritual.html
"Nondeterminism means never having to say you're wrong."  -- Unknown

pgsql-performance by date:

Previous
From: "Victor de Buen (Bayes)"
Date:
Subject: Re: Atomic access to large arrays
Next
From: Merlin Moncure
Date:
Subject: Re: Atomic access to large arrays