Oooh, 3 days in a row! …and all the ladies scream ‘Oh Nic, you are just a mental stud!’ Ha! That’s right, I’m packing heat and not afraid to use it! Seriously though…I know sarcasm when I hear it ;) On a completely random tangent: Does anyone ever click on the links I include or let the mouse hover over those links? You just might want too :)

Work…so yeah, work. I am off to the great cultural haven of TN on Sunday again. I’m much less excited about being gone for a week this time, I’m having way too much fun here in MT. To be honest, it really feels like summer just got started and that I’m starting to move on where I want life to go in the next year or two. Now we are going to hijack it for a week (for work!!). It really isn’t that bad, but the summer keeps seeming shorter and shorter. The project I am working on is still pretty cool, and I am enjoying the limited number of things on my horizon for it. It turns out you can work and have a life at the same time, and life can be fun! Anyways..back to work (yes, I’m just a bit distracted today…) So, my role in this project has turned a bit of a corner, and I am going to blather on and geek out for a bit on it, mainly to get my thoughts in order..

So, it started off as a pretty typical ’support’ engineering consultant. Mr Consultant, please get systems X, Y, and Q up and running, test them, and see where we stand for stability and performance. Ha I say! No problem, whack-a-doodle & presto! This went all fine, thankfully. The XT3 is a big sexy pile of hardware, but no where near the X1E. It behaved pretty well, and I’m looking forward to seeing that 15GB/s I/O throughput number in the next month or two. Shhh, but that just might be a world record…sssssh! So, now that we have demonstrated decent application benefits on a small scale (one app saw 10X improvement), we are moving forward with a broader scope of small scale testing to show the benefits of putting Lustre into production on Jaguar. Jaguar reboots so often, in the 1-2 times a day range, and takes so long to reboot, that adding a pile of service nodes to the equation tends to slow down the whole reboot process. To wrap it all up — basically we need to show the gain that offsets the pain of adding these service nodes and Lustre startup to the reboot cycle.

Now, onto the fun bits. I suppose I should have registed this earlier, but the nature of my consulting to ORNL needs to change to be more of a ‘Project Manager’ rather than a support engineer. What is the difference you ask ? Well, first off, I need to be much more visible and start producing the usual status, progress and update documentation. This really is a critical piece, as I can do all sorts of organization and get people to do the ‘real work’, but if there is no representation and dissemination of that info, well, nobody but me will know that. Not good. I am OK at managing people, and dealing with the logistics and technical details of a large plan like this, but staying focused and on track with it all is not my strong suit. Some of you may remember last summer…ugh. Hrm, well really this is much better, as I know what I am driving towards, and the projects have a fairly defined scope and plan already. There will be a bit of a struggle to change the dynamics and perception of those at ORNL that I am the ‘go-to guy’ for the Lustre storage there, especially since I am working remotely 3/4 of the time. The changes are pretty significant to make, but can be described pretty simply as ‘taking ownership’. It’s mine, and I will cry if I want to!

So, all in all, not that tough to really do, just needed some thinking through for what it all means. Ahh, right on cue, there is a problem to diagnose on the test system…excellent :)