And thank you for taking the time, Rod. The linktionary has a pretty good definition, though I don't know if it counts as "textbook". Same for Wikipedia http://www.linktionary.com/f/fault_tolerance.html http://en.wikipedia.org/wiki/Fault-tolerant_system
Anyway, we need to limit the scope of this document. I think we're only interested in clusters that provide high-availability, that are, to use your terms, completely or partially transparent. Also, while there could be clusters with m members, where n, such that 1<=n<=m, are active, for the purposes of the discussion it is enough to assume that we're dealing with two types of high-availability, mostly transparent clusters: - Those where exactly one does IKE and IPsec, and the rest just synchronize state, and - Those where more than one member are doing IKE and IPsec, all the time synchronizing state. This taxonomy is needed, because some of the problems that affect one cluster type do not affect the other. So I propose the following terms, and these are not for generic clusters providing generic services to the generic Internet, but only for the purposes of this work item in this working group. - For the cluster with just one member doing IKE and IPsec, I propose "hot-standby cluster" - For the cluster with several members doing IKE and IPsec, I propose to keep "load-sharing cluster" Is this fine with everyone? Yoav On Mar 23, 2010, at 1:51 PM, Rodney Van Meter wrote: >> >> I think this is a really nice taxonomy and think it might be useful >> to integrate it nearly as-is into the HA document. >> > > Go for it. I can't promise more help (I'm in workload-shedding rather > than workload-accreting mode right now), but if it's useful, it was > worth an hour of my time to write up. > > There *must* be a standard textbook/reference on FT/HA, but I'm out of > date. > > --Rod _______________________________________________ IPsec mailing list IPsec@ietf.org https://www.ietf.org/mailman/listinfo/ipsec
