Re: ΄πΈ΄: Horizontally scaling a database

I don't think you can do this easily. However, if you must, you can divy up data across multiple servers. On each server you will have processes waiting to work through AQ.

On master you take the query, send it to all servers, they all execute the exact same query, but since they all have distinct subsets of data, they return the dataset to master. Master then processes the dataset to produce the final results.

You can do this using AQs and propogation techniques, but it won't be easy.

So, then again why are you bent upon horizontal scaling? if you can afford multiple cheap servers, why can't you but some upgradable blades, attach them to a x-serv RAID and do a proper clustered database. Granted you will run into licensing issues, but then you can get beefier blades and reduce cpu count.

Oh and I like the performance metric you put out. 2 seconds to return a row from a billion row table. You think that is realistic?

rjamya

On 3/16/07, Sandeep Dubey <dubey.sandeep_at_gmail.com> wrote:
>
> Shareplex is replication tool. I am not looking for how to replicate
> data. I am looking for something like "partitioning a table across
> servers"
>
> Thanks
>
> Sandeep
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

-- 
-----
Best regards
RJamya
-------------------------------------------

--
http://www.freelists.org/webpage/oracle-l