[mapguide-users] MG Scale Out - isolating MG Repository out of MG server process

[Jackie Ng]

Yes, replication of data between the various mgserver instances is an
unsolved problem should you take the load-balanced approach.

Maybe an architectural workaround for the replication problem would be:

Have each server-tier become its own server/web tier. You serve out to
client users/applications through the load balanced web tier, but
author/load site repository data through each server's respective web tier.
This way, you can directly communicate against a specific server's resource
service APIs when loading data/resources, but from the client's perspective
it shouldn't matter which mgserver the load balancer directed them, whatever
resource they request should be available on all the available mgserver
instances.

This idea hinges on whether an mgserver instance is able to handle and
accept requests from 2 or more different web tiers. Never been tried before
to my knowledge. It would be interesting to try out.

- Jackie



--
View this message in context: http://osgeo-org.1560.x6.nabble.com/MG-Scale-Out-isolating-MG-Repository-out-of-MG-server-process-tp5207766p5207844.html
Sent from the MapGuide Users mailing list archive at Nabble.com.