On Fri, May 29, 2009 at 3:31 PM, <[email protected]> wrote: > on the other hand, for server uptime outside of maintinance, you can buy > commodity hardware that will run for years without any failures (your > upgrades are during the allowed maintinance), but you can have another box > of the same type that fails in a month, so you still need to have the > service be HA. if things go well you drasticly exceed your SLA, but if you > removed the redundancy you run a very real risk of drasticly failing to meet > your SLA with a single incident. how do you account for this?
I wouldn't remove the redundancy completely but getting the right amount of redundancy is more economical. Example 1: You have 2 redundant servers so that one can go down and the other will take over. Do you both also need expense RAID controllers? Would software mirroring be sufficient and cheaper? Example 2: You build your web clusters with dual-loadbalancers in front of 100 machines. Two clusters like this can handle all your traffic. You build a 3rd cluster so that any one can be down for upgrades or repairs. You get 99.9% uptime. Then someone suggests building a 4th cluster so that if one is down for repairs AND one fails unexpectedly you are still up. Is that economical? Could you just do upgrades when there is less traffic (late at night or weekends) to avoid the 4th cluster? Alternatively, sometimes one can permit slower performance during a failure. Maybe you have 2 internet connections, but one is cheaper and slower and only used when the other is down, or it is the same speed but billing is set up differently to assume lower utilization. Tom _______________________________________________ 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/
