Consider the following failover services issues when planning your new cell:
· Will your cell be failsoft or failsafe?
· Will you dedicate core services to single machines to isolate and contain effects of a system failure and minimize one service failure affecting another service?
· How will you replicate core services to ensure service availability to the cell's user population? Can the users tolerate service interruptions? Will you replicate the entire namespace and individual directories of names on separate machines? Which server will have the master core services; which will contain the copies?
· Will you replicate applications to ensure service availability to the cell's user population? Can the users tolerate service interruptions?
· Will your cell use other failover mechanisms such as multiprocessors, disk shadowing, uninterruptable power sources, and so on?
· How will you test your cell's failover procedures after initial cell creation and thereafter? How often will testing be done? By whom? Will you test locally as well as across LAN and WAN environments?
· How will your cell's failover procedures be communicated? Who will document them? How often will the procedures be updated?