Support any upgrade activity with zero downtime

We support a large global Spotfire Consumer user base, most of whom don't know that they're using Spotfire (they just know they use 'such and such dashboard'). As such, we need the ability to apply any changes to our environment without user impact, partly due to the impact on operations and partly due to the complexity of communicating any planned downtime (a message saying "Spotfire will be down next week" wouldn't mean much to most users).

With the advent of 7.5 it's now possible to apply hot fixes to web player without downtime, but pretty much every other operation requires some form of downtime. We've created complex processes to limit downtime during major upgrades, but it would be far more preferable if the platform where built with the principle that it should be 100% available during any planned activity.

Some specific items:

• Allow multiple servers in a cluster to run different versions (currently as soon as you update the database, older versions are no longer supported).
• Allow rolling updates by enabling sharing of user session data across clusters. Our load balancers could route users to another server while a server is down, but currently their session information (and all their work) is lost.
• Allow the Node Manager service to be updated without restarting all services on that node.
• Allow Analyst to connect to different versions of the server. Currently you can connect to multiple different hot fix levels, but not different major releases.

I'm not asking that all versions of server/node/analyst be compatible with all others. I think it's perfectly reasonable to put a limit of 2-3 major releases on backward compatibility - so for example an upgrade of 7.5 to 7.8 would be supported without downtime whereas a larger leap may require upgrades in steps (e.g. 7.5 -> 7.7, 7.7 -> 7.10).