Re: Indirect synchronization setup with no synchronizers on servers

David:

First off let me say that you have been very generous with your time
and detailed explanations in all my posts and it is very much
appreciated. I know you have in some cases said the same thing over
and over again and that can be frustrating.

Part of the issue is terminology as you note, my lack of knowledge on
this issue, and I think the other part is the restrictions I'm forced
to operate under. For example, I may be able to get the jet
synchronizer installed on "workstations" but I doubt they will allow
Replica Manager to be installed at all. TSI applications are out of
the question under all scenarios (third-party).

If I get the opportunity to implement indirect then I'm not going to
trust any LAN, except the LAN where the file server is that stores my
application files. That particular LAN is highly reliable and very
fast. But all others will assume indirect replication. You're
probably thinking that this is strange given my earlier stance on
direct synchronization if indirect is not available. I don't mind
putting the coding work in if indirect is available to me.

I'm going to study these posts and then setup some indirect
synchronization on my two laptops and desktop. At this point I need to
be better able to communicate what I need and I need more experience
with indirect synchronization.

Thanks.


David W. Fenton wrote:
rdemyan@xxxxxxxxxxx wrote in
news:1166826432.774590.125920@xxxxxxxxxxxxxxxxxxxxxxxxxxx:

In another thread that is getting very long, jacksonmacd suggested
a method for using indirect synchronization in a scenario where
remote PCs have synchronizers but the server does not.

There is no difference between a synchronizer operating on replicas
and dropboxes stored on its local LAN (e.g., on a server) or on the
same machine where the synchronizer is running.

OK. Let's pursue that suggestion. General setup: Four LANs that
are connected via WANs. One LAN server is the main server for the
entire setup and is where I've been told I can place all of may
application files.

I foresee problems here. You can't use a single server, as all the
remote machines won't be local to it. You need four separate servers
(using the generic term, where a "server" is a machine that offers
shares to be accessed across the network).

Assume Jet Synchronizer is installed on user PCs. Assume that it
CANNOT be installed on any server. Assume further that we have to
use one of the user's remote PCs as the indirect sync hub for each
LAN.

There is a problem with your terminology. "Server" means to you "the
machine that IT calls a Server, the one that is a file server, an
Exchange Server, a web server, etc." while what you mean is "all the
machines that can share files."

I don't understand the messaging in this setup. The problem with
any PC on the LAN is that you can't see it. You can only see the
servers.

The remote machine doesn't *need* to see the machine running the
synchronizer. It only needs to be able to see that machine's
dropbox, which can be stored on the server (and not on the machine
running the synchronizer). The machine running the synchronizer
doesn't have to host the managed replicas being synched by the
synchronizer -- those can be stored anywhere on the synchronizer's
local LAN.

So does this mean that:

1) there needs to be a dropbox on a server contained within each
LAN.

YES.

Each user of the app will have a synchronizer. That synchronizer
will send messages to the common dropbox located on the LAN
server.

Users local to a LAN don't need to use indirect replication. They
can synch directly with the hub on the server because they are
connected by a LAN. Indirect is only needed to synch between
servers. This is the multiple star topology, where each LAN is a
small star with direct replication and the centers of each star
synch with each other (or with one central hub) over the WAN via
indirect replication.

Or is
the suggestion to use direct synchronization for computers within
each LAN.

YES.

Why bother with indirect when direct is safe?

If I can get IT to approve synchronizers, then I would want to
use indirect synchronization. After the beating I've taken about
possibly using direct syncing why would I want to use it between
remote PCs within the same LAN,

Uh, what? What do you mean by "remote PCs within the same LAN"? If
they are connected by LAN, they aren't remote to the server on that
LAN. If they *aren't* connected by LAN, then they aren't on that LAN
at all, and might as well synch with a single hub somewhere else.

when indirect syncing is a viable option.

Use indirect only when you MUST, i.e., when you don't have a LAN
connection to synch across.

2) The designated user's machine on each LAN is the one that will
initiate the syncing of the dropbox message(s) with the replica
farm on the respective LAN server. However, presumably this user
is also using my app. Can his/her synchronizer handle both tasks
(i.e. sending messages from the PC to the server dropbox and
syncing the messages in the server common dropbox with the replica
farm on the LAN server?). If so, why can't any user's synchronizer
do this. After all, all user's synchronizers are sending messages
to the common dropbox on the LAN server (unless the idea was to
use direct syncing within each LAN).

That's a very complicated set of questions. I'm not sure I can
answer, and I'm certain it's relevant once you grasp the other
things I've posted in this newsgroup today.

However, I *think* the correct answer to the last question is that
each replica's messages are specific to that replica. Remember, the
synchronizer drops files in the dropbox on the *other* end of the
connection, not in its own dropbox. If a workstation initiates an
indirect synch with a server, then the workstation drops messages
into the server's dropbox. The server's synchronizer pings that
dropbox every N seconds (dunno exactly how often) and when it finds
messages, it processes them. It then drops messages into the
workstation's dropbox that tell the workstation what data to send.
Then the workstation's synchronizer drops the appropriate messages
into the server's dropbox.

It wouldn't really be different if the synch were initiated at the
server, as the first set of messages exchanged just establishes
whether a synch is necessary (by evaluating generations for adds,
updates and deletes), and that will need to be determined between
any particular pair of replicas.

3) Then we need to synchronize the replica farms on each
individual LAN with the replica farm on the main server for this
entire setup (the central server for this setup of LANs). How is
this initiated?

You set up a synch schedule on the server where the replica farm
replicas are stored.

I think I can see how the topology works on this, but I'm unclear
on how the synchronizers will work since there are none on the LAN
servers.

What matters is *not* where the synchronizers are running. A
synchronizer can run on the server at the center of one of the LANs
or on one of the workstations attached to that LAN. The dropbox and
the managed replicas can be on any machine that can serve files.

Actually, big WHOOPS there. I've said that several times. It's not
true. For an indirect synch, the managed replica has to be local to
the file system of the machine running the synchronizer. Or, it has
to be in a hidden share (one that ends in $).

It's the dropbox that can be anywhere on the local LAN. The managed
replica(s) that serve as the indirect replication hub must be on the
same machine as the synchronizer. If they aren't, then you run the
risk of getting a direct synch because the remote machines can see
the replica on your end.

--
David W. Fenton http://www.dfenton.com/
usenet at dfenton dot com http://www.dfenton.com/DFA/

.

Relevant Pages

  • Re: Test to see if another remote is syncing with the hub replica
    ... But don't you still need to install the Jet Synchronizer on the server. ... making replica backups, restoring replica backups (in case the server ...
    (microsoft.public.access.replication)
  • Indirect synchronization setup with no synchronizers on servers
    ... PCs have synchronizers but the server does not. ... Assume Jet Synchronizer is installed on user PCs. ... one of the user's remote PCs as the indirect sync hub for each LAN. ...
    (microsoft.public.access.replication)
  • Re: Indirect synchronization setup with no synchronizers on servers
    ... remote PCs have synchronizers but the server does not. ... There is no difference between a synchronizer operating on replicas ... One LAN server is the main server for the ... use one of the user's remote PCs as the indirect sync hub for each ...
    (microsoft.public.access.replication)
  • Re: Please ease my terror
    ... >> from a hub residing on the server under these circumstances. ... > from the existing replica set, so this wouldn't be the end of the ... > synchronizer to the remote replicas. ...
    (microsoft.public.access.replication)
  • Re: Do this plan sound feasible?
    ... However the connection is too slow to run the DB over. ... Terminal Server on one end of the connection. ... create another Replica, which I will put on a network drive in the ... Kaplan's TSI Synchronizer Object to do indirect synchronization. ...
    (microsoft.public.access.replication)