Replication and Deleting Rows

Hello,
I have a large table with over 20 million rows.
It is replicated accross a 4 subscribers using Pull subscription. One of the
subscribers is also a Distributor.
We are running some processed to "clean up" the data in this table by
issuing a lot of DELETE commands (using Primary Key as the where clause)
against the large table. The DELETE commands are enclosed in BEGIN TRAN -
COMIT TRAN batches of 20 each.
The deletion works extremely fast on the Publisher.
However, it takes very very long time to apply the very same commands on all
of the subscribers. The subscribers also have the PRIMARY KEY index that is
on the Publisher.
We need to delete a few million rows from the large table using this
technique. However, it makes the Subscribers very slow and unresponsive.
Any suggestions?
Thanks,
Arsen
Consider doing this in even smaller batches. If that is not possible, and
transactional consistency is not really of importance, then you could use a
parameter called MaxCmdsInTran for logreader (introduced in SQL Server 2000
SP1 I believe). This breaks huge transactions into smaller chunks of
specified size, while storing the commands in distribution database.
There is another approach, that I would recommend. Consider using 'stored
procedure execution' replication. With this method, you keep a copy of the
archiving stored procedure on all subscribers. When you run this procedure
on the publisher, the distribution agent will call this sp on all
subscribers, instead of sending millions of DELETEs over the network. See
SQL Server Books Online for more information on this.
Vyas, MVP (SQL Server)
SQL Server Articles and Code Samples @. http://vyaskn.tripod.com/
"Arsen V." <arsen@.community.nospam> wrote in message
news:%23hpi4tKoEHA.2612@.TK2MSFTNGP15.phx.gbl...
> Hello,
> I have a large table with over 20 million rows.
> It is replicated accross a 4 subscribers using Pull subscription. One of
the
> subscribers is also a Distributor.
> We are running some processed to "clean up" the data in this table by
> issuing a lot of DELETE commands (using Primary Key as the where clause)
> against the large table. The DELETE commands are enclosed in BEGIN TRAN -
> COMIT TRAN batches of 20 each.
> The deletion works extremely fast on the Publisher.
> However, it takes very very long time to apply the very same commands on
all
> of the subscribers. The subscribers also have the PRIMARY KEY index that
is
> on the Publisher.
> We need to delete a few million rows from the large table using this
> technique. However, it makes the Subscribers very slow and unresponsive.
> Any suggestions?
> Thanks,
> Arsen
>