One of the biggest issues with ANY costing model is choosing the cost
which you are to recover.   The choices can make or break a particular
platform, this is especially true when there will be an attempt to
compare the "total" cost of two different architectures.

As an example, I worked for a firm that had a Mainframe mostly
datacenter for many years.  In the post 9/11 need to replace
inaccessible distributed sever farms (they weren't considered
datacenters), the mainframe datacenter was used.  The datacenter was
built in the days when a M/F took up a large footprint, dasd took up a
humongous footprint  and now there was more than half of the datacenter
empty.    Upon migration of these distributed servers, the M/F manager's
cost started to increase, as the 8 chillers that were shutdown years
before, were restarted, and additional ones added.  As the power
consumption skyrocketed, his bill did also.  Now this shop was a "full"
cost recovery shop, at least on the M/F.  The datacenter mgr realized
his cost per cpu minute was about to go through the roof because the
billing only took into account the capital cost of the machine, the cost
of  M/F software, the applications use of dasd,  and spread the total
cost of the datacenter (including staffing)  evenly over all z/OS and
z/VM cpu secs, this without changing the M/F at all.  So he  told the
accountants to calculate the increased costs of the facility as part of
the distributed cost, and he realized they were now using 50,000 sqft of
very expensive floor space so he charged that to the distributed and in
the end, with the reduction in sq ft charges, and savings in staffing (4
fewer server locations), the z/OS and z/VM cpu second charge went
down.    The distributed groups went berserk!  They had never been
charged for floor space, nor had they been charged for power or cooling.
these charges were covered in the overhead numbers of their business
units, since the server locations were in the same spaces as the BU's
they supported.

Now all the platforms saw the real cost of power, cooling, floor space
in addition to cpu,memory,disk, network and the M/F started to look just
a little bit more cost effective, especially when they started to build
out the mandated DR site, at the second M/F datacenter.

This customer did not embrace z/Linux at first (in fact they threw it
out the door once) but has since reintroduced the platform and is
actively expanding it's use.  The change of heart in this case occurred
when a multi-million dollar expansion of the power infrastructure was
going to be needed to implement a new distributed application, in
addition to the application's other costs.

I will agree this is not an exact science, but at least it doesn't have
to be a skewed one either.

Phil



Stewart Thomas J wrote:

Best way to be equitable is via usage based charging, ala cell phone billing. 
First goal is to get a figure for the total costs that you want to recover from 
all your customers and over what period. Let's say it's 100 dollars over 1 year.

Then you figure out some categories to charge back for. Cell phones get minutes 
or text messages. For servers you want things like CPU and memory that are easy 
to gather via z/VM accounting records.

So let's try to recover 25% of our costs by charging for memory. So find out 
how much memory all your servers will be using (lets say 10GB over 1 year) and 
come up with a rate per GB of memory and charge each server some rate per month 
or year. Note that this won't be exact because what you forecast and what you 
really end up with over a time period might be more or less, so you might over 
or under charge. It's not an exact science.

Do the same thing with CPU. Estimate total CPU for all servers for some period 
then divide that into your $75 to get a rate per CPU second. This is hard the 
first couple years until you build some history of how much different servers 
consume. But this is basically what phone companies do to come up with a per 
second rate.

If you end up getting too much money back, you can use that to enhance your 
services, or rebate back to your customers, or lower rates next time around. If 
you undercharge, you raise your prices next cycle.
__________________________________
Tom Stewart
Infrastructure Analyst
John Deere - z/OS Support Services
em: [EMAIL PROTECTED]
__________________________________




-----Original Message-----
From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of Mike Hamner
Sent: Tuesday, July 01, 2008 2:41 PM
To: [email protected]
Subject: Costing Model for Linux-390

I'm new to this group.  We are also a new z/Linux shop and are currently 
bringing up z/Linux.  We are trying to determine a good costing methodology to 
apply to this environment to be able to distribute cost back to our customers 
in an equitable manner.  Does anyone have experience with doing this and have 
you defined a cost model strategy that you would be willing to share?

Michael E. Hamner, CDCP
IT Architect/Engineer
Chief Technology Office
National Business Center
U.S. Department of Inerior
7301 W. Mansfield Ave.
Denver Co. 80235
303-969-6619

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions, send email to 
[EMAIL PROTECTED] with the message: INFO LINUX-390 or visit 
http://www.marist.edu/htbin/wlvindex?LINUX-390

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390




--
'in media stat virtus'
Virtue's in the middle

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to