RE: Saving the System i: Fight Rather Than Switch

------------------------------

message: 4
date: Wed, 13 Dec 2006 11:45:52 -0600
from: "Jones, John \(US\)" <John.Jones@xxxxxxxxxx>
subject: RE: Saving the System i: Fight Rather Than Switch

Brian - Possibly, and I admit we haven't tried to price that 

scenario,

but the numbers say a consolidated box will need 24-30K CPW and 

that's

potentially several 520s.  

What we need is an inexpensive but fast engine for WebSphere App 
Server.Please withhold the oxymoron comments.  What will run WAS 
fast & cheap?

1 big 570? Yes and No.  
1-3 Windows servers + a moderate (existing) 570? Yes and Yes.  
1 moderate 570 + 2-4 520 Express machines? Yes and Maybe.

Here's what we're up against: A Dell PowerEdge 2950 with dual Quad-
coreXeons w/2MB cache per core (Quad core is really 2 dual-cores 
lashedtogether and each dual-core has 4MB cache), 16GB RAM, 
Windows Server 2003 Enterprise, mirrored 73GB disks, and a 3-year 

gold warranty 

has a list of just over $14K.  Probably closer to $12K after our 

discount.

That's around $1500 per core for a complete server with tons of CPU
capacity, adequate disk, RAM, OS license, and a warranty.  RAM may 
be a little shy, I'm not sure, and the 32GB RAM feature is quite 
pricey, so I'd add a second server instead of upping to 32GB RAM.  

That would 

alsoprovides some redundancy.  So for under $25K I'd have 16 
cores, 32GB RAM, and all the trimmings.  Can the iSeries compete 

with that?  

Becauselike it or not, iSeries iNtegration advantages or not, like 
the platform or not, this is what it boils down to.

(What might count under other circumstances but doesn't this time
around: WebSphere App Server license - we have unlimited.  Server
Administration costs - mostly a wash across scenarios.  Data center
impact - another wash as 2 of the Dells take just as much resource as
the additional CEC in a 4/8-way or 1 extra 520.)

That's where point 3 of my hardware points makes sense from IBM.. they 
already have identified one of their systems as loosing ground because 
of cost, identified several area they see as problems and gave a lower 
cost solution to address it.  Was a strategic plan to stop some 
processes from leaving the Mainframe..

"Earilier post - Point 3 "
   3.  The biggest - the processor.  I'm not going as far as Steve to 
say that every processor should run all out.  I would rather suggest 
a 'compromise'.  We are in the process of getting a new mainframe 
(first new one in 15 years).  It's a small BC (Business Class) box.  
When you look at the configurations, you really get ship 4 processors 
and they turn on how much capacity you need.  Sound familiar.  The 
other options are to turn on other processors as FULL engines to run 
things like Java, Linux, and DB2.  That doesn't go against the z/OS 
licensing or any of your third party products.  If your need two 
processors to run you RPG/ILE code and DB2/400, fine but let me run 
the Webshpere or HTTP processes on the other FULL, not OS/400 charge 
processors.
    I realize that sounds like a difficult task, and probably is, but 
the way the got CFINIT sitting there ready to pounce you'd think these 
types of things are possible.