Archive for January 27th, 2010

Jan 27 2010

New Land – 1km square in 1 full region of ScienceSim.com

Published by under SL In General

Good stuff has been happening – and GIS data has been finding its way into OpenSim!

Thanks to the persistence of Kim Smith and with help from John Jainschigg, I made my first-ever live audience presentation in the World2Worlds venue on 2009 12 15. Trying to fit everything into a regular work day, I was fortunate to get support to do the presentation from my desk at work rather than from home–and I indulged in a new digital-USB headset for the occasion. Oddly, the headset seemed invisible on the podium ;^)

Speaking at Smarter Technology, 2009 12 15 -- John Jainschigg built a custom podium several steps tall so that people could see my tiny avatar.

Speaking at Smarter Technology, 2009 12 15 -- John Jainschigg built a custom podium several steps tall so that people could see my tiny avatar. Photo courtesy of Chimera Cosmos

After nearly a year away from pushing OpenSim (corresponding with my first year on a new RL job), I was a bit nervous about having up-to-date stuff to offer the audience. So instead of a slide stack, I presented the machinima of models that I had brought from OpenSim into Second Life.

Lazy of me, perhaps, but the video did help create a certain party atmosphere in the airy auditorium at W2W. It also helped echo my emphasis on a certain use of OpenSim that I’ve been striving towards since October 2007.
But as so often during RL conferences, the very best part of the presentation experience is the (sometimes serendipitous) human connections that take place around it, and this presentation was no exception to that rule.

One audience member was Richard Hackathorn, who described Second Life in the context of urban planning for listeners to Arina Hadich’s Urban Design Podcast. His enthusiasm for my subject matter at the Smarter Technology presentation has led to an upcoming podcast at UrbanDesignPodcast.com (details to follow). Although I haven’t been a regular listener to podcasts, when the subject of the talks are focused on an area of interest, I am impressed by how much information can be conveyed in the time it takes to listen. The podcast seems to have more detail per minute than I can get reading a browser on my phone, certainly more than I can browse while driving, (and far more detail than I seem to be able to convey in a minute spent writing a blog ;^)

Another audience member was Shenlei Winkler of Fashion Research Institute, who is surely among this world’s most prim-prolific individuals. I’ve been reading about the various Shengri La regions since 2008, created by Shenlei in collaboration with IBM researchers. Ironically, I was collaborating with different IBM researchers about the same time, in 2008 03. These days, Shenlei appears very active with several types of support for the IEEE/ACM-hosted, OpenSim-derived grid known as ScienceSim.com. Shenlei very kindly took time to contact me after the Smarter Technology presentation, introduce me (with voice chat) to ScienceSim and its resident researchers, and encourage me to participate in the Science Sim Land Grant program. More details are described in paragraphs below.

Through ScienceSim, I’ve had the pleasure of interacting with Mic Bowman of Intel Labs, the group managing ScienceSm servers. The way that these folks have configured the servers is glorious. I’ve been a fan of running OpenSim on 64-bit Ubuntu with Mono since 2008 07, and the Intel crew have taken it to such another level that I find it astounding. In ScienceSim, it appears that the ODE physics engine runs for as little CPU cost as that which I’d experienced before with the near-trivial basicphysics. It is a rare treat for me to speak with configurators of OpenSim, much less those who strengthen and extend the simulator code!

Since I was firing up the forges to create some LiDAR sculpties for my ScienceSim project, I decided to warm up the works by creating some carefully-scaled terrain for ScienceSim’s Yellowstone 16-region model. Using public terrain data from the US Geological Survey, I processed terrain down to a model that was 1:83 in the x and y dimensions, and 1:55 in the z dimension (vertical scales are often exaggerated in both hominid avatars and terrain to make them more attractive). As a by-product of the production process, I also saved an intermediate bit of data that I had scaled to 1/10.38 in x and y and 1/6.9 in z. Knowing that it would present a worthy challenge to OpenSim server jocks, (the 1:83 model fit into 16 regions, but the 1:10 model would require exactly 1024 regions) I passed it to Mic along with the 1:83 terrain.

My jaw was somewhat slackened when, less than a week later I heard from Mic that much had been done with the
1:10 terrain. The Intel crew had actually had the tenacity to wait while all 1024 regions were brought up on a single processor; shortly afterward they prudently fit the regions onto a single blade, dual quad-core (Xeons?) system with 11 GB of memory and my favorite X86_64 Ubuntu/Mono environment. As if that wasn’t enough, they did this while configured with ODE physics! I’d say their effort was Olympian, but heck, it actually took them less than six days!

It’s been tremendous fun this month watching OpenSim make bold moves in the direction needed to support the sort of civic paraverse / Immersive Connected Experiences that could back-end many aspects of local agency operations. I’ve been looking for ways to get here since 2006 11, but with Intel’s demonstration of 1K regions on 1 dual-quad-core Xeon blade, I won’t look like such a fool scoping out costs for a county that would require 20K regions to build out at nominal 1:1 scale. (By contrast, in 2007 I had estimated that Linden Lab hosting of 575 regions would cost the City of Berkeley upwards of $60,000 per month!)

The Yellowstone work by Mic has been beautifully documented by Shenlei here, and here and here and here. I’m even using one of Shenlei’s Yellowstone sunset shots as my (Ubuntu) desktop background ;^)

More to the point of my post’s title, I have taken some outstanding publicly available LiDAR data and am in process of crafting a 1/3.9-scale model of Marin County’s own Point Reyes Station, a rural community that is situated so close to the San Andreas Fault that they were imaged in excellent detail as part of a scientific investigation into the geomorphic expression of strike-slip faulting. Together with publicly available high-resolution orthophotography, I’ve been able to refine the technique that was used on the 40-region model of the UC Berkeley campus and develop 21-cm-gridded surface model with 10-cm natural color orthophotography. For a preview of the model at 105-cm surface gridding, view this video. Enjoy, and watch for updates soon.

The Point Reyes Station may be trod upon in ScienceSim’s Alaska North region. Hope you get a chance to visit!

No responses yet