Wow, what a weekend! More than 3000 people at the Moscone Centre for the conference. We got to hear all the business unit EVPs, Jonathon and Scott. Andy Bechtolsheim's enthusiasm for the four
product lines his team is working on is incredibly infectious. I can't wait until we see these new products in the field --- you're going to love them.
Jonathon came out in a basketball singlet which --- to me at least --- looked like it was made for somebody a good 1/2 metre taller than he. Jonathon's fairly tall too....
He talked for about 10 minutes to start with (without slideware), and then did 55 minutes of q and a. It is really something to be listening to him when he takes questions, because he really is on top of everything and has an illustration for every point he needs to make. And these are honest
points and illustrations too, which as technically-minded person (and audience) I really appreciate.
John and I presented on the new features that are in Solaris CAT
v4.2 for crash dump analysis. We're currently working on the legal stuff so we can get it onto the Sun Download Centre
(look for the Operating Systems link). The team has written a whole heap more system "sanity checks" and added new options to commands like "cpu" and "rdi" to help with various frequently-performed operations. We are also now able to provide you with the actual function arguments
that a function was called with almost all the time --- this used to be a tedious two+ step operation. We've got support for Solaris 10
as well (if you're running a Solaris Express build that isn't at least s10_72 then please upgrade
, some critical structures changed). My favourite part was being able to demonstrate an alpha (well, maybe pre-alpha) version of Solaris CAT
running live on my laptop which is running a amd64/64bit solaris next nightly build. I think there were a few "oooh"s and "aah"s .... that's something which is planned for our v5.0 release.
Then there was the demo room!
I caught up with Roy and Mark at the SGR booth (I'll write about SGR in another entry; it's our troubleshooting methodology), said hi to the SunLibrary people who were demonstrating Grokker (again, for another entry), had a look at the new products from Andy Bechtolsheim's team and then spent a long, long time with the engineering folks who demonstrated FMA aka Predictive Self-Healing
FMA is the Fault Management Architecture which we've built into Solaris 10 software-wise, and are progressively building into our new hardware. (It's already in the SF280R, the SFv1280 and lots of others). It is a mechanism whereby Solaris can accurately remove marginal components from use by your system. At this point --- as I understand it --- the minimum size component that will removed is a cpu core, or a memory page. I don't think I can quite do justice to the whole architecture and explain how it works, so I won't. You should visit the pages under Predictive Self-Healing
to get a better understanding. Suffice to say that if you are running Solaris 10
on your system, then FMA can do an amazing amount for your system Reliability Availability and Scalability.
In a previous entry I mentioned Sun Preventive Services
, and that team had a booth in the demo room too. It was great to see their evangelism and the little flashy-red heart shaped things they gave away. It was also great to know that one of my colleagues in Sydney is one of the drivers in this service offering.
David Yen's talk about CMT was very, very interesting. I know that in a few years I'm going to be studying cpu design at university. I expect that it will be a few months after our Niagara chip arrives in volume systems, so the lecturers might well not have caught up with what will be the next revolution in cpu design. While I'm thinking about it, one of our Performance and Availability team is already working on figuring out how we will need to re-design, re-think and re-educate the concept of a unix load average. As far as I can determine, Niagara-based systems mean we'll pretty much have to throw out a whole heap of established wisdom and re-invent it. That will take some time!
I got to two breakouts today, one (from one of our German cluster aces) on some of the experiences he has had with SunCluster
and a fairly large OEM/ISV and the other on Solaris 10 performance from the technical lead on that project.
I think it's fair to say that the thing I can say on my blog is this:
If you're going to run a clustered environment, Architect (or design) it right. Implement it right and Maintain it correctly and you will have one of the most solid and reliable environments you can possibly get.
The Solaris 10 performance presentation was given by Bart Smaalders
, a bloke who really, really knows what he is talking about. There are so many perfomance improvements in Solaris 10 that I can't list them all (let alone remember them all). I'd like to say that I've been seeing these improvements myself even on my old nearly-retired duron-650 system and they are only magnified on my laptop. If you aren't running Solaris 10 then you really, really, really should be. I had a good long chat with Bart after his presentation and found out a few more things about other planned changes which are at the scope-out stage right now. Solaris performance is really hotting up.
We finished the conference with Scott answering questions for more than an hour with the sort of frankness, honesty and good humour which are his trademarks. I haven't worked for other companies, so I can't compare with them, but I believe it to be truly rare that a COO/President and CEO/Chairman not only take questions and answer them frankly, but that they and all the business unit EVPs make it a mandatory part of their year to turn up to a Services event like CEC. To me it indicates the worth of our work to Sun and that gives me a really good feeling deep inside.
I've had a blast at CEC meeting people from all over the world, presenting, seeing the new features, products, services and solutions that are going to be released soon, and learning as much as I can to take back to Australia and pass on to my colleagues.
Now I've only got another few days here in the SF Bay Area --- another conference, StorageAces --- and then I get to go home. Woohoo!