That logic is completely false and you contradict yourself. Allowing for multiple points of failure does not mean that something is less reliable as you have described. It means that if/when one fails, the other will still be available.
Using your example of a power supply lasting 10 years, that would translate to 2 failures in 10 years, not 1 failure in 5 years. I think you understand the concepts, as your "best" solution is to have multiple points of failure with failover using CARP. Jim O'Donald -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of chefren Sent: Wednesday, August 03, 2005 12:47 PM To: misc@openbsd.org Subject: Re: Ammunition needed to defend OpenBSD/pf On 08/03/05 19:25, Shawn K. Quinn wrote: > You mean having the DSL router and modem be in the same physical box, > thus introducing a single point of failure? That's a huge minus. ??? You would prefer a milion boxes for each individual transistor or logic gate? Two boxes have two CPU's, two power supplies etc in the same production line and the total is thus =less= reliable than a single box solution. If the power supplies are equal and each statistically break down once in ten years in total that becomes once in five years. By the way, the same applies for RAID, more concurrent harddisks definitely means =more often broken drives=. Although the system won't break down if well designed, you still have to do more repairs. Five drives instead of one: Five times as much disks to repair/replace. Good is often to have a spare, pre installed(!), "DSL router and modem", better is to have a concurrent and tested(!) backup channel. Best is to have a working backup channel: CARP! +++chefren