On Fri, 19 Jun 2009, Brad Knowles wrote: > [email protected] wrote: > >> I claim that it's not possible for a two servers to have the same state. or >> at least not with acceptable performance. > > A lot depends on how you achieve that "same state" condition. If you've got > shared storage, then you don't have to migrate the storage. Even if you do > have to migrate the storage, do your storage update requirements on a > per-second basis overwhelm the bandwidth you've got to the hot-standby > machine?
I agree, but point out that you are ignoring memory updates, and with network servers you have memory things that need to be updates (TCP sequence numbers for example) > I mean, mainframes have been doing this sort of thing for twenty or thirty > years. You can have a bomb literally take out an entire datacenter, and the > backup mainframe can take over without a noticeable hiccup. This is because > both machines are kept in lock-step synchronization, and nothing gets > returned to the caller until the remote system has responded that the > operation is complete. > > Expensive, yes. But certainly do-able. And something that could be > replicated by more modern systems. first off, the mainframes were mostly doing batch stuff, and secondly the applications we designed to do checkpoints periodicly and restart from the last checkpoint. also a key phrase in your statement is 'a noticable hiccup' I absolutly agree that you can keep two systems 100% in sync, but you can only do this if you never fo any changes at a rate higher than the bandwidth between the machines (and add the nessasary delays to all changes to accomodate the latency between the machines) however, I don't believe that anyone (other than 'man rated' space and submarine systems are willing to accept the performance hit that this causes. David Lang _______________________________________________ Discuss mailing list [email protected] http://lopsa.org/cgi-bin/mailman/listinfo/discuss This list provided by the League of Professional System Administrators http://lopsa.org/
