On 05/23/2012 07:11 AM, Herouth Maoz wrote:
> Hi guys,
>
> I'm interested in a solution that will allow our customers to run reports - which may involve complicated queries -
ondata which is as up-to-date as possible.
>
> One thing I don't want to do is to let the reporting system connect to the production database. I want the indexes in
productionto be limited to what production needs, and not add indexes that are required for reports, for instance. And
basically,I don't want a customer to run a complicated report and degrade the performance of my production system.
>
> A replication solution is not very good, either, because of course I can't define indexes differently, I don't want
*all*transactions in all tables to be sent, and also, because I may want to cross reference data from different
systems.So ideally, I want to have a reporting database, where specific tables (or maybe even just specific columns)
fromvarious databases are collected, and have a reporting tool connect to this database. But I want to push the data
intointo that database as close to real time as possible.
>
> The most important data I am currently considering are two tables which have an average of 7,600 transactions per
hour(standard deviation 10,000, maximum in May is 62,000 transactions per hour). There may be similar pairs of tables
collectedfrom more than one database.
>
> I assume this is not an uncommon scenario. What solutions would you recommend?
Hi Herouth,
I would have a look at slony. It is a trigger based replication system
that allows you to replicate only the tables you define and you can have
different indexing on the slave. The only requirement is that each
table you want to replicate has the same primary key or unique
constraint on the master and slave. Other than that, you can index the
tables on the slave however you want.
Cheers!
- Chris
