This section addresses Req. D1.2-7.19-DynaUpd.
For continued availability of the system, Zero-Downtime-Update (ZDTU) technology SHOULD be implemented through out. If horizontal scaling path and failure recovery have been implemented, then ZDTU can be implemented easily by taking out of farm one server at a time and updating it. Downside of this approach is that the farm will temporarily be in an inconsistent state.
If consistency of the farm is at all times a requirement, no easy ZDTU approach exists. One approach is to bring up new "hot standbys" along side of the old configuration and then do instantaneous switch. As the switch over is less than 1 second, this could be considered ZDTU.
Never-the-less, as TAS3 is business process driven and as business processes can take long time to complete (if human interaction is required, this could easily mean days or weeks), thus consistent ZDTU is infeasible in practise and the business process modelling should explicitly foresee handling of upgrade situations, i.e. how old processes are handled after the general upgrade.