Re: i5 Youngsters

On 1/15/08, Jones, John (US) <John.Jones@xxxxxxxxxx> wrote:

Hardware quality may be similar but it is not the same. X86-class
servers are still engineered to a price point; POWER servers are
engineered to a desired performance/reliability standard. Price is a
factor but is not the overriding concern.

I disagree. This might've been true in the past, but i doubt it
applies today. Just open an x3650 and a Model 515. One of the two has
a well thought out interior. The other one doesn't.

Disks are essentially the same (though the i5 uses previous-gen
hardware, U320 instead of SAS), and doesn't have 2.5" disks.

PSUs are both from some factory in China or Taiwan, of the same quality level.

Internal bus is PCI-X (previous gen) in the System or PCI-E
(current-gen) in the x3650.

Of course, a low end machine like a System x3105 is different.

Your hardware sample size is arguably small. I've worked with over 300

It is.

AS/400s through System i systems over the years and my current employer
alone has over 1000 x86 servers. I've seen AS/400s run for years with
failed hard drives and have seen x86 servers literally have their
motherboards melt due to PSU failures. The x86 servers are "decent",
BTW, with brand-names like Dell, IBM, & Compaq.

I've seen a model 150 go up on flames, but i do not consider that to
be because of cheap hardware used by IBM ;)

As to software quality, specifically the OS, I'll only say this: We
rarely _have_ to apply i5/OS PTFs to fix a problem. Rarely means less
than annual. We do apply them periodically (a couple of times a year)
just to get current. There is no i5/OS "Patch Tuesday" because there
doesn't need to be one.

That's mostly because IBM has an "interesting" policy when it comes to
Security. IBM just ignores the problem forever, while Microsoft
releases a patch a few months later. Both are bad, but IBM is just
worse.

Just look at the version of BIND that ships with i5/OS - it's horribly outdated.

And look how many PTFs are constantly released - a clear indicator
that there ARE problems in i5/OS (which makes sense, it's made by
people after all).

In all other aspects the new environments is radically more complex, has
more components, more points of failure, costs more to operate, requires
more time to administer, has more difficult problem solving (involves
multiple teams v. just one), has more vendors to work with should issues

Why do you split that up onto multiple teams? Now, i'm not that
knowledgeful about IT in big american companies, but to me that
doesn't make alot of sense.

In theory it's more scalable and it probably is, but so is the iSeries
as we're in a 570 and are using nowhere near 16 CPUs. So where you'd
just "add a few front end servers" I'd just call IBM and get a license
to unlock a CPU.

A single, slow sub 2 Ghz CPU for a ridiculous price. Or buy a new
server with current-gen hardware, slowly phasing out the older front
end servers.

Where it is definitely better is the capital cost of scaling, however
higher operating costs will overcome that over time.

Maybe. But it will be able to provide a lot better uptime than an i5
IF done by people with the proper knowledge.