Simon Josefsson schrieb am Wed, Jan 09, 2002 at 10:45:52PM +0100: * [EMAIL PROTECTED] writes: * * > We also built HA for each mailstore, such that it are in fact two * > systems clustered by the kimberlite software mounting a shared * > RAID in a failover situation. See * > * > http://oss.missioncriticallinux.com/projects/kimberlite/ * > * > for details. * * Did you considered more decentralized fail-over? I'm thinking about * having synchronized IMAP servers present at physically separate * locations. Any thoughts on this?
No. The project we did had much priority on being able to support a very large userbase without having to buy hardware for 1k bucks and more. With this restriction, we had to give performance considera- tions 100% attention. Systems on geographically separate locations that aim to be synchronised are a very good idea because this very much enhances reliability. Anyway, I do not know of any distributed FS (say, intermezzo or gfs) that is - fast enough - stable enough to faciliate the task at hand in a manner that makes the system work for 1M users and is at the same time not consuming very large amounts of money (i.e. some companies like IBM may have working concepts for this, but I would not dare to ask for the price). Upon inexpensivety: Maybe there is another idea possible. Kimberlite does sort of a parallel mount to a RAID on a shared SCSI bus. Maybe there is a way here to achieve low-level distance. If you have a bus that could be transported over some length - say: fiberchannel over ATM? - there would be no need to bother what the OS does on it. The other think to keep in mind is that you have to provide services on the same IP adresses to do HA - on distant locations you may get problems with your routing if some machines switch service IPs on and off being in separate networks. But here as well this might be solved with low-level concepts like IP-over-ATM or VPN-like setups. Regards, - Birger
