[AccessD] Replication - how can I tell if indirect occurring

John Skolits askolits at ot.com
Wed Jul 16 08:53:54 CDT 2003


" The difference was dramatic, so much so
that I think it is much superior to the classic local-FE + server-BE
layout."

For now, I think I will try your model instead of a single BE. I may then
put some code on the server that handles the synchronizing when I tell it
to. I'll see how this works.

Thanks!

John



-----Original Message-----
From: accessd-bounces at databaseadvisors.com
[mailto:accessd-bounces at databaseadvisors.com]On Behalf Of Arthur Fuller
Sent: Wednesday, July 16, 2003 9:09 AM
To: 'Access Developers discussion and problem solving'
Subject: RE: [AccessD] Replication - how can I tell if indirect
occurring


I could be wrong about the 5 minute synch. It was a few years back that we
did the replication stuff. After a while we upsized to SQL 2000 and I
haven't had a need for replication since.

At the peak of the setup, there were 4 branch offices, each with a server
and a synchronizer, and a replica on each local PC in each branch. The
server at each branch synchronized all local PCs. The server at HQ
synchronized with the servers only at each branch. Only 4 synchronizers
handled approximately 70 PCs, without a glitch in over 6 months of constant
use. That may seem like very few, having to do lots of synchronizing, but
when you think about it carefully and perform a few calculations, what
actually must be transmitted? In most cases, a few longs, a date or two, a
text column or two per row added by a user. How many rows can someone add in
15 minutes? Not many -- let's say they're really fast and can enter 2
complete transactions per minute (i.e. two rows in the master table, a few
more in the detail table(s)). Add to that the replication columns' data and
you still don't have much to synchronize. It was a lot faster than the
standard FE-BE.

I have a lot of respect for Mike Kaplan, but even more respect for empirical
evidence :-)

Incidentally, the reason each PC had its own replica is that this approach
eliminated lots of net traffic, since all the reads to populate combo-boxes
etc. came from the local hard disk. The difference was dramatic, so much so
that I think it is much superior to the classic local-FE + server-BE layout.

Arthur

-----Original Message-----
From: accessd-bounces at databaseadvisors.com
[mailto:accessd-bounces at databaseadvisors.com] On Behalf Of John Skolits
Sent: July 15, 2003 9:25 PM
To: Access Developers discussion and problem solving
Subject: RE: [AccessD] Replication - how can I tell if indirect occurring


Arthur,

I guess my reasoning for using indirect with the VPN connection is based on
what I've read and what has been recommended by other newsgroups and MS.
This includes the Guru - Mike Kaplan.  Everyone pretty much stated that
direct synching over a VPN provides alot of network traffic and if the VPN
connection drops out, database corruption is almost a sure thing.

I have no plans replicating the FE, just the BE. My original idea was to
have one BE on each server. (1 for each location Phili and Cleveland)Then
only replicate the BE when needed between the 2 sites. Also, the
synchronizer seems to only allow 15 minute increments, so I'm not sure how
you did 5.


Your statement:
"A synchronizer handled branch-level replication."
Do you mean you had a server that had one BE to many FE's?

Your statement:
"The server at HQ handled server replication."

So there was on BE sitting at HQ, that replicated with only the 4 "branch
level"synchronizer?

Therefore, based on your description, you only had 4 synchronizers running
with 70 replicas?



-----Original Message-----
From: accessd-bounces at databaseadvisors.com
[mailto:accessd-bounces at databaseadvisors.com]On Behalf Of Arthur Fuller
Sent: Tuesday, July 15, 2003 2:33 PM
To: 'Access Developers discussion and problem solving'
Subject: RE: [AccessD] Replication - how can I tell if indirect occurring


First of all, let me apologize for the slow response. The last two-three
days there has been an area-wide net-outage here. (Class action suit anyone?
I notice the bill doesn't change when it's the ISP's fault.)

Why do you want indirect synchs, particularly when the servers can see one
another? I don't see the logic here. Indirect is designed for situations
where the BE's cannot always see each other, for example a laptop in a
distant hotel used by a travelling salesperson, who connects using the net
and uses FTP for replication. If the replicas can see each other and you
don't want replication to happen automatically, don't use the synchronizer.
Instead do a manual (unmanaged) replication. /Tools/Replication/Synchronize
Now, IIRC. Further, I still don't understand your need for three
synchronizers. One can do it all, AFIAK. Of course, maybe I'm missing
something important here.

Finally, I have never experimented with replicating an FE, only a BE. But
for a long time (until we moved to SQL 2000), I replicated the BE's all over
the place without issues. At one point, there were over 70 replicas,
distributed across 4 branch offices connected through a vpn. Each office had
a server and a collection of PCs. A synchronizer handled branch-level
replication. The server at HQ handled server replication. Both processes
occurred every 5 minutes, so the longest it would take anyone anywhere to
see changes made elsewhere was 10 minutes. It never failed once in over 6
months of use.

I could be wrong about all the above. I found a scheme that worked
beautifully and stuck with it. There may be better solutions.

Arthur

-----Original Message-----
From: accessd-bounces at databaseadvisors.com
[mailto:accessd-bounces at databaseadvisors.com] On Behalf Of John Skolits
Sent: July 14, 2003 12:52 PM
To: Access Developers discussion and problem solving
Subject: RE: [AccessD] Replication - how can I tell if indirect occurring


I will have one replica here (my officer) and the other 2 on remote servers.
All servers connected to each other via VPN. Can't do direct because I've
read it's a bad idea to do that on WAN.

I think I'm slowly figuring it out except, I've read where:

"To prevent direct synchronization from occurring, make sure the replica is
not stored in a shared folder; "

I want to have the replicas on 3 servers all with synchronizers running with
indirect Synchs. Then use those replicas as backends to various frontend
applications. The FEs will be linked to the replica in their perspective
local server.

But, in order to link I have to share the database folders. The white paper
indicated, if I shared the folder, then a direct sync would occur (implied:
Even over a VPN).

So it looks like I can't have a FE/BE design on replicas that run with
indirect synching.

What do you think?

John

_______________________________________________
AccessD mailing list
AccessD at databaseadvisors.com
http://databaseadvisors.com/mailman/listinfo/accessd
Website: http://www.databaseadvisors.com


_______________________________________________
AccessD mailing list
AccessD at databaseadvisors.com
http://databaseadvisors.com/mailman/listinfo/accessd
Website: http://www.databaseadvisors.com

_______________________________________________
AccessD mailing list
AccessD at databaseadvisors.com
http://databaseadvisors.com/mailman/listinfo/accessd
Website: http://www.databaseadvisors.com




More information about the AccessD mailing list