Alan Hodgson wrote:

>On Tue, Jan 04, 2005 at 11:11:49PM +0100, KubaTyszko wrote:
>  
>
>>connection
>>Reply-To: kuba at lbl.pl
>>
>>hi all.
>>i have two servers placed about 300KM from each other.
>>both have 256kbit/s connection.
>>one will be master and another will be slave.
>>(master has some services like squid redirector that for each http
>>request does select from database.
>>will putting slony into work wslow down this squid redirector ?
>>here is exactly what i mean: there is database i wish to replicate,
>>master host has redirector that needs about 50-100 request to one of
>>tables per second.
>>updates / inserts to database are not often - they occur about every 15
>>minutes.
>>will installing slony slow down the redirector or not.
>>    
>>
>
>Slony slows down inserts and updates somewhat, because a trigger runs
>that inserts a row into a log table for each updated or inserted row in
>each replicated table.  There is also some overhead for reading this data
>out to execute on the slave, but not a lot.
>
>It does not slow reads at all.  If you have very little write traffic
>you probably won't notice any impact.
>  
>
There's a possible _slight_ exception to the "does not slow reads at 
all" claim.

If it proves necessary to _add_ a unique column to the table in order to 
do replication, then that adds a bit (well, 64 bits) of data to each row 
in the table.  That'll make reads a tiny bit slower.

But that is of course not a "normal" case; if there is a well-designed 
schema, there will already be a suitable primary key, in which case 
you're exactly right.