[lug] More Server Problems

[George Sexton]

> -----Original Message-----
> From: lug-bounces at lug.boulder.co.us 
> [mailto:lug-bounces at lug.boulder.co.us] On Behalf Of Sean Reifschneider
> Sent: Monday, March 06, 2006 12:44 AM
> To: Boulder (Colorado) Linux Users Group -- General Mailing List
> Subject: Re: [lug] More Server Problems
> 


> >calendar, I'm thinking I'll be able to run something like 
> 500 virtual hosts
> >a machine with 2GB of RAM. If I can get 64 bit boxes with 16 
> GB of ram then
> >I would be able to run something like 4000 or so customers 
> on one machine.
> 
> Those rarely make sense unless you REALLY need the big-arse 
> box.  Which it
> doesn't sound like you do, you aren't working with any single 16GB
> data-sets, etc...

No. The individual data set are really small. I just hope to have a lot of
them.

> 
> These sorts of boxes usually start at $15k and go up pretty 
> quickly from
> there.  If it's your only box, what are you going to do for 
> your clients to
> ensure availability?  Spent the big dough on the 4 hour 
> on-site support and
> field the calls the 4, 6, 8, or more hours until they've 
> fixed it?  Buy
> another cold spare box, doubling the price?

That's a pretty good point.

> 
> For that same $15k you can get 17 1U lower-end machines, with 
> a total of
> 16GB or even 32GB of total memory, *AND* have a cold spare, 
> plus reduce

The problem then is that I have 16 times more machines to update and manage.
Additionally, the uptime formula is (if I remember right) the (Number of
Boxes * SquareRoot(Probability of Failure))^2. My own experince tells me
that if I have 16 machines with drive mirroring, 2-4 of those mirrored
drives are going to die in the first year.

> sales dictate, instead of having to spend the $15k or more up 
> front and
> then find the clients.

I don't think there was every any chance of that <g>.



George Sexton
MH Software, Inc.
http://www.mhsoftware.com/
Voice: 303 438 9585