Log in

AGU Day 1 & 2 - The Mad Schemes of Dr. Tectonic [entries|archive|friends|userinfo]

[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

AGU Day 1 & 2 [Dec. 13th, 2016|10:16 pm]
It's only been two days of conference, but I feel like I've been here for a week. Being away from home and routine makes everything seem to take up a much longer span of time.

I started Monday with a talk that made me think that I had to rework the core of my bias correction method, but once I left the room I realized, wait, no I can't replace the transform with that approach, because it depends on already knowing the thing that I use the transform for. *Whew!*

After wandering the posters a bit, I sat in an overcrowded and really interesting session about cloud computing. And then another one about it that afternoon. And another one today. And it's got me convinced that we need to take a new tack on some of the projects back home, to the point that I'm gonna try to get some experiments going on my own. So that alone was probably worth attending for.

Since I got out of the pre-lunch session a bit before the lunchtime rush, I zipped over to the Metreon and got lunch from Sorabol (Korean) and took it back to my room.

We had a (productive) side meeting from 5-6:30 that night for one of our collaborations. I bowed out of dinner afterwards to go to the MIT alumni reception, but it was at a new venue this year and turned out to be a total bust. Quincy was wiped and had to cancel, and it was in a tiny room with dismal refreshments. There was one table of food, with a pedestrian cheese platter, vegetable crudites and hummus, and three types of sliders. Oh, and a dessert table you couldn't get within five feet of. I bailed after maybe 10 minutes, and got dinner instead at a place called Lemonade, which is a sort of salad + hot dish fast-casual setup, emphasizing fresh and local. I was pretty pleased with what I had, honestly.

I've slept poorly the last two nights, due to foreign bed and overly-squishy pillows. I called down for a couple firm pillows and I'm hoping I'll do better tonight. But 8 am came really early this morning. Along with all the cloud computing talks, there was a good session on data formats. And I was able to say hello to Kristy (one of my grad school cohort) at the posters afterwards.

I had a very good lunch today meeting up with Jason Funt for lunch at the Whole Foods a few blocks south of here. Hooray for meeting internet friends in person! We got along well, and I wish we'd had more time.

I had a big gap in my schedule after lunch, so I rested up and actually used the hotel jacuzzi for a bit to loosen up the soreness of bad sleep and lots of walking. Dinner tonight from Loving Hut in the food court at Westfield. It's miscellaneous Asian cuisine and vegan, but more importantly, the food court was nearby and it was the one that was most appealing when I went wandering by it.

[User Picture]From: ocschwar
2016-12-14 04:25 pm (UTC)
So what is the deal with climate scientists and the cloud? Have you guys given up on speccing your own hardware and fine tuning your code to run on it?
(Reply) (Thread)
[User Picture]From: dr_tectonic
2016-12-14 06:13 pm (UTC)
We totally do that when it comes to running the models, but it's a lot harder to do that with analysis. We do lots of analyses that involve gigantic piles of stupidly parallel computation, but they're generally one-off computations, not things we're going to do over and over.

Demand-wise, it's super-spiky. I might want 10x the crunch power of our entire supercomputer, but only for about 3 seconds, and only a few times a month. That's hard to accommodate with dedicated hardware that needs a good cost-to-usage ratio to justify its existence, but ideal for elastic computing on the margins of other people's server farms.
(Reply) (Parent) (Thread)