Re: i5 Youngsters

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

I don't doubt that the layout of the 515 is not service-friendly but not
having seen one I'd have to wonder if the layout is designed for better
airflow or makes sense from some other strategic standpoint.

The machine is short, like a small pizza box. Maybe to make dual-use
as a tower easier.

performance-tuned firmware. I hope we can skip 2.5" platter-based
drives and go straight to SSDs, either 2.5" or 3.5".

Well, the whole SSD thing *is* going to be interesting, but i doubt
that we'll see SSDs in mainstream servers in the next two or three
years.

that do better with larger size and/or better airflow. I also think IBM
tends to size them a little larger than necessary so they aren't as
stressed if the chassis is maxed out.

That is usually the case with most "better" server hardware.

There's nothing wrong with PCI-X beyond it not being popular.
Performance-wise we aren't maxing it out (yet). Not being popular will

Transfers from an IOA Cache can surely benefit from PCI-E attachment.

drive up adapter expense some, but they can engineer bridge chips easily
enough to adapt PCI-E cards to PCI-X. That said, POWER6 systems have
both PCI-E and Infiniband in addition to PCI-X. See section 2.1.2 of
RedBook 5052.

Yes, i know. But there are no POWER6 based smaller machines.

Sure, i5/OS has problems and PTFs are released pretty much daily, but
how often do you have to actually load a PTF to address a problem you're
having? It's exceedingly rare for most shops to have to do that. Also,
the OS is huge and includes the database and all of the other bits we
love so PTFs cover a lot of territory.

They do. And the only thing about PTFs that i really hate is that you
can't automate them. (which would be really beneficial for very small
shops that only get help if somethings broken).

I've had several cases this year in which i needed PTFs. Web Query is
proudly leading the list with 5 PMRs and uncountable PTFs. Next comes
optical library handling.

Oh, and when we upgraded our test machine to V5R4 a few days after GA,
things weren't uninteresting either.

Bind and other OSS-type apps do tend to be behind the latest. But as
long as there isn't a critical bug in the old version I don't really see
anything wrong with that.

Older versions of bind are susceptible to a variety of DNS poisoning attacks.

FWIW the box in question isn't sub-2GHz; it's a POWER5+ with 48GB RAM
and 3TB DASD (RAIDed capacity). Our other i5 is a POWER5 and will
probably go 5+ or 6 sometime this year.

The fastest POWER5+ CPUs are 2.2 Ghz and dual core. They're way behind
Intel's current offerings.