"Gordon Burditt" <gordonb.0jlgv@burditt.org> wrote in message
news:12490m9bvsplifc@corp.supernews.com...
> >I have a client that desires a new application for use at several stores
> >in
>>different cities. He wants "real time" access to all store data and the
>>ability to continue to run locally even if the internet connection goes
>>down.
>>
>>It occurs to me that a local MySQL instance should definitely be
>>in each store to continue work if the net is down for any reason. But,
>>what
>>is the best way to synchronize/share data among all of the stores?
>>
>>From what I have read on the MySQL site, you need MySQL Cluster to do
>>synchronous replication, but it won't work over a network connection less
>>than 100 mb.
>
> That's not the only form of replication. Plain old MySQL can do
> replication although it doesn't have all the features of the cluster.
> You might have to manually switch masters if the net goes down.
> Depending on the query volume and how up to date the slaves have to
> be, a 1200bps modem connection might be sufficient.
>
> My employer tried a mail server setup with MySQL Cluster, using
> MySQL to maintain info on where the mailbox was and user options,
> across several machines. It didn't work very well. We'd get
> inconsistent data and storage node crashes that would essentially
> take out the whole setup. Eventually we used conventional replication,
> with changes going to the master and each machine using its local
> copy for reads. Perhaps MySQL Cluster has improved since then.
>
> What is your required data setup? Each store is master of its own
> data and the head office also wants a real-time (read-only?) copy
> of it? Or does the head office need to make changes, and stores
> need to share each other's data? I don't know how to do conventional
> replication with, for example, the head office machine slaving one
> database off of each store master.

The concept that they came up with was that they wanted real-time data
access to do store-to-store inventory lookups and transfers and to track the
store activities from the central office.

They also want to share customer data (in case a customer visits another
store or in the event that a customer has not finished paying for
merchandise and is behind on payments).

They also want all store data backup up in case of a catastrophic event at a
store (like theft of IT resources or fire).

They also want a local store of (at least) that store's data - or (if
possible) all store data until the connection to the central database was
lost.

And, let's not forget a way to post back to a central database (if one is
used) any local transactions done at a disconnected store once the
connection to the group or central database is restored.

Simple, huh?

JH