I'm seeing more and more of that in the field, Larry. Back in the day, you could conceivably replace a smaller AS/400 or iSeries with much cheaper hardware and get comparable performance. I'm not saying it was a good decision, but it was at least understandable. But today even the smaller IBM i boxes have a price/performance ratio that makes it nearly impossible to match with commodity hardware.

I'm dealing with a similar situation where just a tiny piece of the LOB is being replaced a Unix/Oracle combination. The requirements? Seven virtual servers on two physical boxes to provide redundancy that isn't even at the reliability level of the original box. Two servers for Oracle, two servers for the application, two servers for the communications infrastructure, and a 7th server for MQ Series to communicate with the host. Note that with all that we STILL have a single point of failure in the MQ Series server.

And this, by the way, doesn't include the additional hardware and software costs for a separate QA machine, because you don't want to roll changes out to your production server.

And with all this, the system is exceedingly fragile. Oracle just got back to us because the (redundant) production server went belly up with zero load. (Luckily, we're only in setup phase not actual production.) They found a bug in their failover code. It took nearly a week to isolate it.

If ever the phrase penny wise and pound foolish applied, this would be it.



What seems to be missing through the thread here is how few servers
you need when running POWER especially with IBM i. Applications
integrate so much more easily when running on the same platform than
when there is a little here and a little there. Recently one of my
customers was told by a vendor that a piece of their application which
formerly ran right next to their primary LOB application was now going
to be running on two windows servers and an AIX server. It would now
need to extract data from their LOB app, move mash process manipulate
exchange convert translate and finally return it to the LOB app in the
form where it could be imported. The argument was that those three
additional pieces were each 'Best of Breed' and had been integrated to
provide the best possible solution. Even assuming that's a true
statement how much better is it than the previous solution and how much
money will that either save for the customer or how much additional
sales will it enable? Will it pay for the three servers? Three operating
system licenses? A copy of Oracle? Far more complicated backups?
Increased power and cooling required? And the 800 lb gorilla, managing
and troubleshooting of this far more complicated solution. Tightly
integrated the original application was just 'there' and worked. Now
when something doesn't work it's hunting around for what's broken, what
server has an error, low disk space, a service that didn't start, a
thousand things. Worse neither of the two new operating systems were
ones the customer had skills with so even more cost is incurred for
those skills.

This thread ...


Return to Archive home page | Return to MIDRANGE.COM home page