File replication via sattelite

Hello.
I need to do replication via sattelite communication. This means: Very high cost (7$/minute), low datarate (64 kbit), not reliable and long delay (up to 4 seconds from a datapackage is sent until an ACK is returned.
I would prefere to do this offline- that is by transfering a file with replication data.
The datavolume shuld be kept to the lowest possible.
A high degree of data partioning is possible in the application.
High degree of autonomy is necessary- only replication once a day.
This suggests merge replication.
But is it possible to that filebased: That is, transfer a file (via e-mail) and then do the merge at the subscriber totally offline?
I would appriciate if anyone could lead me in the correct direction on this. (Or stop me before I reach the edge..)
The short answer is no. All replication types use rpc or port 1433 to
communicate with each other. Going to something as synchronous as a file
transfer will not work.
Note that Exchange essentially does what you are looking for. Exchange
servers send email (IIRC) to each other telling them what to do.
Hilary Cotter
Looking for a book on SQL Server replication?
http://www.nwsu.com/0974973602.html
"otto b" <otto b@.discussions.microsoft.com> wrote in message
news:88E097FC-6CB1-4E37-B0B0-A92260E6FA75@.microsoft.com...
> Hello.
> I need to do replication via sattelite communication. This means: Very
high cost (7$/minute), low datarate (64 kbit), not reliable and long delay
(up to 4 seconds from a datapackage is sent until an ACK is returned.
> I would prefere to do this offline- that is by transfering a file with
replication data.
> The datavolume shuld be kept to the lowest possible.
> A high degree of data partioning is possible in the application.
> High degree of autonomy is necessary- only replication once a day.
> This suggests merge replication.
> But is it possible to that filebased: That is, transfer a file (via
e-mail) and then do the merge at the subscriber totally offline?
> I would appriciate if anyone could lead me in the correct direction on
this. (Or stop me before I reach the edge..)
|||Partitioning of data according to subscriber is most easily achieved using
merge and the -HOSTNAME parameter. It is also possible in transactional
replication but more difficult to maintain: you use Transformable
Subscriptions and define a different DTS package for each subscriber.
There is no filebased method of transferring the replicated data, apart from
the initial snapshot files of course, so using email won't work.
Minimizing the data volume is more difficult to determine. Mostly
transactional is less data, but if you had 1000 updates of a single record,
that would be 1000 transactions in transactional replication and just a
single record in merge, so recommending the method 'depends'. You could
measure network traffic in a test environment and compare transactional and
merge for your scenario.
HTH,
Paul Ibison