Re: Repository services and budget

[Kent Crispin <kent@songbird.com>]

Jeff Williams allegedly said:
> 
> Kent,
[...]
> > In this model, the performance issue isn't redundancy and
> > ultra-reliability -- it's the computational cost of doing public key
> > operations.  These operations are O(frequency of domain creation) --
> > for a registry creating under a thousand domains a day a high end
> > pentium machine should be more than adequate.  Something that could
> > run in a closet at ISI.
> 
>   I suppose you have taken into account Hardware maint. and drive
> failures
> as well?  How do you perpose to handle that with one machine? 

I wouldn't.  I would keep a duplicate spare.  Two pentium machines.  
Maybe $10000.

> Or is the
> that registry just down for that period of time?  

You can't reserve or add any new names to DNS during the period the 
machine is down.  Something like two hours, worst case.  In the case 
of a catastrophic failure, where the disk is lost, you wait an hour 
or whatever for the registrars to propagate their DNS updates.  If 
DNS is down as well you have a pretty serious problem.

>And is the domains
> on a spicific drive not avalible when a drive goes bad?

It doesn't keep any record of domains -- only pending updates and 
reservations.  The information concerning domains is in DNS.  If DNS 
goes down for the zone, then you have a serious problem.  But that is 
almost impossible -- there should be multiple secondary DNS servers 
for the zone.

-- 
Kent Crispin				"No reason to get excited",
kent@songbird.com,kc@llnl.gov		the thief he kindly spoke...
PGP fingerprint:   5A 16 DA 04 31 33 40 1E  87 DA 29 02 97 A3 46 2F