I think you've about hit it on the head. If DB io is the bottleneck then
you might want to look at IBM i. But, as far as I can tell it would
probably have to be a 2x improvment just to even consider it because you
could probably throw more lintel hardware at it and make up the IO
difference for less than going with IBM i.
Thanks
Bryce Martin
Programmer/Analyst I
570-546-4777
James Rich <james@xxxxxxxxxxx>
Sent by: midrange-l-bounces@xxxxxxxxxxxx
03/08/2011 05:27 PM
Please respond to
Midrange Systems Technical Discussion <midrange-l@xxxxxxxxxxxx>
To
Midrange Systems Technical Discussion <midrange-l@xxxxxxxxxxxx>
cc
Subject
Re: i as a dedicated "LAMP" server (was Re: IMHO article on database
performance)
On Tue, 8 Mar 2011, Mike Wills wrote:
Okay, so this has been bouncing in my head for a while and have been
meaning
to ask.
Given this research. What is stopping a web startup from buying an IBM i
(a
relatively small one with plenty of growth) then breaking it out into
two or
more partitions. One partition would be for the database, the others
would
be load balanced as web servers with PHP.
I think that the primary obstacles are, in order:
1. price
2. knowledge of/skills with the i
3. awareness
Here are the questions that come to mind in something like this:
Could a single IBM i take more hits per second than a two-server LAMP
setup?
My experience is that the database on IBM i can do more transactions than
MySQL. As for the rest of the system, I haven't really tested much, but
in general it seems like linux is pretty smokin' fast. But make no
mistake about it, when it comes to database IBM i wins easily.
What about more LAMP servers?
I don't know, but I would guess adding more machines would add enough
complexity that I would immediately recommend an IBM i (if the database
were the bottleneck).
What are the cost differences for the initial purchase of the two
options?
Big. The smallest IBM i that I have been able to spec comes in at around
$10,000 when all is said and done. The last server class linux machine I
ordered for a customer was $3,000 from Dell. There are reasons for the
cost difference, but the difference is hard to ignore.
What about expansion costs?
I think this would be hard to quantify. That said, I think the IBM i
tends to require expansion less often.
James Rich
if you want to understand why that is, there are many good books on
the design of operating systems. please pass them along to redmond
when you're done reading them :)
- Paul Davis on ardour-dev
midrange.com
