Re: Alpha remembrance day

Main, Kerry wrote:
-----Original Message-----
From: David Mathog [mailto:mathog@xxxxxxxxxxx] Conversely there is a compelling reason to move off of pretty
much all of the HP OS's since they are all slowly sinking into the
sunset while tied to the Itanic. Migrating to commodity hardware
and software offers a better, cheaper future.

The HW platform is one of the smallest parts of an IT budget. IT staff,
training, business logic compatibility, new functionality QA and
regulatory testing, application licenses and support, DC costs etc are
the biggies. Changing the underlying HW is one thing, but also changing
the OS as well is a 800lb gorilla to address.

You're overstating the difficulty of migration to another OS/platform
versus the problems of not being able to obtain software for the current
platform at all.


And by commodity hardware and software, do you mean adopting OS
platforms that *average* 7-20 security patches per month?

Presumably you mean windows? Actually I was thinking more along the lines of the x86 variants of Solaris or Linux. These too have security
patches of course, whether or not they actually need to be installed
depends a lot on what the server in question is being used for. That
is, a hole in libjpeg is not a serious security hole if there are
no interactive users. My experience with these two OS's has been that
I can generally install all security changes (via RPM) on Linux without
requiring a reboot. The cumulative patches for Solaris often do require
a reboot.

However the argument is irrelevant since the OS which doesn't require
these patches (VMS) typically no longer has the required software available and so is not an option in any case. Or to put it another way, a steel strongbox will keep your secrets safe but you can't run SAP on it.


How would you
price the cost associated with QA ing and testing all these security
patches each month with all of the Customer apps running on these
platforms?

Insignificant compared to not being able to run the apps at all. You are of course talking about customer developed apps and I am talking
about commercially supplied ones.

Most Customers are concerned about their apps and as long as there is a
support plan in place for them, the HW costs are usually a much smaller
consideration.

By "most" you mean "customers with very special needs and deep pockets" whereas by "most" I mean "customers who use off the shelf software".


Case in point - why are Solaris Customers not migrating en masse to
Solaris on x86?

There is no compelling reason to do so since the software they need exists on Sparc/Solaris and so the migration cost isn't justified. In
that instance all they would be doing would be changing hardware without
any significant change in any other costs. That isn't the case for
getting the heck off of the Itanic, where software availability is extremely limited.

(FYI we recently looked into migrating one Oracle Sparc/Solaris
server to Oracle Opteron/Linux when the application that ran on top
of Oracle finally was supported on Linux. The deciding factor
against the move was that the trade in value on the Sparc server
wasn't enough to buy a comparable Linux system. Had we been setting
up the system now, with $$$ to buy new hardware, it would have been on Linux. Of course this is for Biology research and not banking, so we
don't have either the security or up time requirements you assume
are typical. Definitely not the deep pockets.)

Regards,

David Mathog
.

Relevant Pages