Failover on six /18 "subnets"?

[sthaug at nethelp.no]

> >I'm a little worried by the time DHCP will need to initialize the leases 
> >database and I don't want these large ranges to interfere with our 
> >smaller subnets already in place.
> 
> This is governed by the size of the pools, not the size of the subnets - the server builds a hashed table with an entry for every possible lease it can offer. If your pools are the full /18 subnets, then I make that around 100k addresses to keep track of - and there have been reports of slow startup times when you get to this sort of scale. You wil need a fair bit of memory, and fast disks to manage that - you'll also need to address other factors, such as ensuring that logs are configured as asynchronous.

I think the reports of slow startup times may be a bit exaggerated. I
am running a failover pair with around 100k leases on hardware which
is by no means state of the art (~6 year old Dell 1850) *but* it has a
hardware RAID subsystem with battery backed cache - and the logs are
indeed asynchronous. A restart of the server takes around 20 seconds.

Steinar Haug, Nethelp consulting, sthaug at nethelp.no