Darb Dabney’s OpenSim Blog

I’ve been in a bit of a rut the past couple of days, feeling doubt about which way to proceed with configuring the OpenSim side of the UC Berkeley campus 1.024:1 sim. For the first time since I started setting up OpenSim test servers back in October 2007, I was uncertain of my ability to make it work with this project. I rolled back to 0.4, 0.5.0, 0.5.1, and the trunk that worked a couple of days ago would run only 32 regions well, and even at that would stop working, without any use, by morning. All my effort was going into testing out various ways of retreating from the leading edge. In an activity like OpenSim, that’s not a fun place to toil!  Now, after the sim sits quietly through the night, I can teleport from my landing zone in the far SWly region to the far NEly region, and get there pronto.  Plus the 40 Regions are barely consuming 1% of the CPUs.

Realizing that a good 48 hours had passed, one of the things I tried tonight was a fresh grab of the trunk, and that really turned things around for me. With OpenSim 0.5.4_4272 I have the same rocket-fast launch, zippy association of terrain with regions, and I can actually teleport into regions that haven’t resolved their terrain without finding my av hung up. That was all good. Then, I started moving around the ERDAS Imagine data that will be stamped into terrain megaprims, and I was reminded that I’d gone to all the trouble of resampling both terrain and orthoimage for 40 regions, and my diced file naming conventions were already dependent on that entire set of 5 x 8 regions. So rather than fire up the process for making sculptie bump-maps, I went back to the 0.5.4_4272 build, shut it down and went after my region configuration–willing to give it another try at 40 regions. While I was at it I generalized my PHP region XML configurator.

Just for the sake of enjoyment, here are a few views, including nice Windlight night shots. Compared to two days ago, one can notice more of the hill at LBNL, with the fairly intricate grading for service roads and laboratory buildings quite evident in the scene.

Enjoy!

Somehow, things are harder at full scale than they were just a few months ago doing 1:4 work in OpenSim 0.4 for all of Berkeley. The new OpenSim 0.5.4 has some aspects, like initial launch, that are blazingly fast. Still, I struggled with what surely acts like lazy instantiation in the new code. It launches like a rocket, but when I touch it with a client, lots of the same old slogging starts to happen. Very much most significantly for my extensive and largely static work, the current 0.5.4 will not configure stably for more than 35 regions on one (dual-core) server, even with BasicPhysics.

So, after fighting with it for three days, rolling back toward 0.5.0, and finally flopping down to tonight’s SVN trunk, I have reached a spatial compromise that can be shared. I really wrung my heart out trying to pare down the 40-region design posted last week into an “essential” 35 or fewer regions, balancing the virtues of demonstrating 1:1 scale paraverse work in a relevant way to students, Cal faculty, and City government interests. After hours of exploring options, I threw out the northernmost eight regions, to leave a 4×8 array of 32 regions that retain (somewhat selfishly) the Martin Luther King Jr. Civic Center Building, as well as the California Memorial Stadium (for Hayward Fault interests), the Greek Theater, and the Berkeley Community Theater (for musical memories’ sake).

The terrain is of the finest quality for use in OpenSim. I took Alameda County LiDAR-based bare-earth terrain in triangulated irregular network (TIN) form, and using the ArcGIS 9.2 3D Analyst extension, ground out a 30-cm posting interval grid version, in WGS84 UTM 10 north meters, NAVD88 Geoid2003 meters, for processing in Leica Geosystems ERDAS Imagine 9.1 to produce diced tiles that were flipped and exported to raw single-precision floating point, byte-swapped (Motorola style) raw binary terrain tiles that OpenSim so readily digests. To implement my chosen 1.024:1 well-tempered scale, each “1-meter” terrain sample in the sim was actually 976.5 mm in sample interval, simply taking 256 samples over 250 meters of real-world terrain for each sim X- and Y- axis. Thus far, I have neglected to scale the Z in the same way, but when I get to stamping out the terrain megaprims, and need to rescale the terrain for best fit, I’ll rescale.

Here’s two views of the sim running on a dual-core 3.4 GHz server (Ubuntu 7.10 / Mono). One view from the Synchotron building area of Lawrence Berkeley National Lab, at the northeasterly corner of the sim, looking southwesterly.

The other view shows how incredibly detailed and appropriate the LiDAR terrain is when used in 1:1 scale OpenSim application. The view is from near the SEly corner of the sim, and shows the bare-earth expression of the California Memorial stadium, the Greek Theater, and Bancroft Ave running westerly, as visible by its smooth, crowned road grade and gutters, all plainly visible in the terrain.

Just a note to describe more fun sailing with the new Windlight viewer, and the Wee Tiny Tako 3.2 (that’s a version, not its length. The actual craft is sporting a 2.25-meter waterline!

For my Dad’s birthday this week, I took a sail to the Northern Continent.

Along the way I sailed through ANWR
Finally, I met some unpleasantly private waters that were poorly marked, and my craft reverted to my Inventory, leaving me sitting on a coral reef. Just before this mishap, I enjoyed some fresh winds up to 12 knots northerly and boat speeds up to 7 knots.

Plenty of fun!

This post has little to do with OpenSim, except to share the enjoyment of the new viewer that works well enough with sculpted terrain megaprims in OpenSim. This evening was simple diversion, a novel experience of sailing in a Wee Tiny Flying Taco, learning how to sail it better, and navigating the ancient inland seas from Gualala through Rosedale and Kapor to the timeless shores of DaBoom. Virtual weather was very pleasant with a westerly 7 knot breeze. I moored off Stanford’s southern shore and strolled a bit, and got to see some square-rigged pirate ships up close. The tiny Taco fits well under bridges and as always, Second Life from tiny eyes seems bigger and perhaps more wondrous.

One annoyance, for me, was that the new viewer seems to have altered the rules for focus-and spin when mousing. While learning how to build detailed interiors of structures, I really got used to the click-to-focus, followed by an Alt-click to rotate point-of-view around that clicked focus point. Now that seems lost and I miss it very much. It matters not too much with sailing, but the way it works now would seem crippling to me when it came time to build fancy structures. So for that, maybe OpenSim folks will want to keep the 1.19.0.5 installer fairly close at hand!

I’ve been intrigued by the press release brought to my attention about IBM Virtual Data Centers
There’s plenty to check out with a new 0.5 version of opensim, although when I built it on the Core2 machine here I did not experience a difference from 0.4 that stood out immediately. (That’s fine by me!)

I’ve got new terrain to try out: a 30cm grid posting based on a LiDAR survey from Alameda County Public Works, plus a research-grade LiDAR survey for USGS GeoEarth Scope on the Hayward Fault. Might be time to consider a 2X or 3X model just hollywood pokerpoker sexi gratisholdem poker downloadgioco streep pokergiocare a poker on linegioca pokerpoker in lineapoker no onlinetornei poker on lineplay poker,play poker online,play wize pokercasino tropez bonus codemigliori bonus casinovideo poker on line,giochi on line video poker,video poker on line gratisgiochi keno inlinearegole della roulettegiochi black jackvideo poker online gratisjack black in lineasoftware roulettecasino tropezforum casino on linecasino venezia on linegiocare alla roulettevendita video pokerinternet gamblingroulette online gratisslots machine gratistrucchi video pokercasino gioca,casino game,casino gamingquestionario casino on netsistemi per rouletteprofessional video pokergioco della roulettekeno gratiscasino poker gratiscasino italia gratiseuro casinoil gioco della roulettevideo poker online gratis,video poker gratis,giochi gratis video pokerslots casinoadvanced video poker,giochi video poker,video pokergioco di roulettecasino online certificatikeno inlineavideo slotsweb casinolive poker game,poker game,play poker gamelive poker gamepoker tipholdem poker to see how well one might display crown of road, gutters, catchmens, and storm drains. Hey, with a physics engine, they might catch marbles and model drainage.

Such waves of goodness that it wasn’t possible to blog it all.  Here’s where things have gotten in the past 2 weeks:

Beyond the provided OpenSim command terrain load-tile,  there has been another world.  The great thing about load-tile is that it takes a single terrain binary (in IEEE f32 format) and loads it into a bunch of sims, one terrain sample per square meter.  The challenge is that it’s very Windows-like in its use of system resources, which means that little objects spawn until the cows come home, and never give back their memory.  Solution: more rigorous scripting in the ERDAS Imagine side of things, and cooking up the terrain tiles myself.  The twist, more like a flip, here is that the good folks who coded “read in the terrain” used Cartesian Y=up, X=right, while the age-old standard for raw raster drops down like NTSC fields in a TV picture, Y=down, X=right.

So when I diced up the terrain and transformed the dices into flipped binary, they would load OK.  There were some horrific moments while I figured this out on a bunch of sims, and the world looked like a tectonic Cuisinart had just passed through.  Anyway–with a load of little scripts, and code-to-make-scripts, I have reached 100 regions and have an earlier 64 sims’ worth of Mt. Tam terrain loaded.  Still on 1 GHz and basic physics for now.

-=Darb

Whee! The OpenSim experience grows more wondrous each week. My humble trash-to-testing server, 1 GHz Celeron, 1.5 GB memory, Ubuntu Gutsy Gibbon, is now running at capacity: 49 standalone OpenSim regions, seamlessly covered in 1:4.57-scale real-life terrain, a single user and just the first couple of prims.

The scale works out to a real-life terrain square almost 8.2 kilometers on a side.

The very first couple of prims have gone in to mimic the Gardner Fire Lookout on top of East Peak.

This feels like a bona fide breakthrough for me, knowing that prototypes of Berkeley could be set up, perhaps 49 regions to a test box, on just 10 servers for starters. That would be supporting a 1:1 scale, 500-region model of the City. Much to think about!

Silence is not becoming of a little blogging avatar. I’ve posted here and there, nagging others and such over the past 10 weeks or so. I’ve even missed the 9 millionth SL account somewhere along the line. But despite my silence in this space, it’s not like I’ve been sitting on my spherical posterior!

In August I brought together a few bits of technology, tested out a Vista-based budget PC with a little ATI Radeon X1500 upgrade to its graphics. Lo! and behold, somehow the SL viewer worked OK under Vista. Annoyingly, the MS OS trolls rated the native ATI driver at a lower “Windows Experience” rating than the default MS graphics card driver. But the Pentium Duo box was fairly stomping for around $550 with the upgrade. Later in the month (OK, maybe a couple of days later) I wedged the box into luggage and got it set up about 10 zones easterly of SL time.

The cool part was being set up in a medium-sized eastern European city, with cable modem, trucking around SL and having perfectly nominal voice conversations with known avatars. I may have been many thousands of RL miles displaced from home, but I had to say so to have anyone notice. In fact, they were probably a bit skeptical about my claims (like many people might be the first time they get an overseas Skype call.) Try as I might, there wasn’t any difference in the experience that I could tell between using voice-enabled SL in the SF Bay area, and using it in Romania. SL might eventually turn out to be a pleasant supplement to webcam conversations.

Technology-wise I’ve also gained a bunch of media knowledge in these quiet weeks. Thanks to the good folks at Berkeley Community Media, I copped 24 hours of training in field production and Apple Final Cut Pro; for someone like me who has almost enough undergrad and graduate Documentary Film Studies classes to get a minor, it is wonderfully refreshing to see non-linear editing in action. Sure, I’ve got a copy of Adobe Premiere Elements in the home lab running on a PIII/800 Coppermine, and I keep a Pentium D running XP Media Center edition to get some DV capture action happening. But compared to a Dual-Xeon 3 GHz G5 Mac with 4 GB of memory and Final Cut Pro, it’s kids’ stuff that I have at home.

Add some FRAPS action and the SL script Filming Path (from Geuis Dassin and Nand Nerd), and I feel equipped to exploit the Berkurodam build as a sound stage. Now if only I had talent to match my technology…

I did set up a burning car crash in Berkurodam and then neglected to clean it up for several days. Apparently some of my neighbors were displeased, or a strolling Linden took exception to the air pollution that was released by the fire, and (embarrassingly) the mess was cleaned up by them and returned to me. After that, I’m determined that any carnage left on the Berkurodam set will be cleaned up shortly after filming.

Also, in another mainland area, I’ve built a machinima studio of moderately grand proportions, and had some dialog with not-so-near neighbors who suggested ways of lowering my albedo, particularly at night. With very little effort, I blissfully got with the Goth program and have had some fun. Now the studio is a one-acre tower with a shroud of darkness on the outside, and chroma key blue glowing on the inside.

Again this semester, I’m very glad to say that there is a student group has been willing to set up a term project in Second Life. There’s no more space in Gualala, but there should be plenty in the machinima studio for the time being. I look forward to their construction!

Finally, before I fade into RL dreamland, I want to echo my appreciation for Orange Montagne’s book Metaverse Manifesto that I read over the Summer. It’s great! While I haven’t heard from Mr. O directly, I was pleased to post a review of the book on Amazon where I got my copy and recognized today that he’s read the review.

Here we are again at another million accounts. Something like five weeks each for the sixth and seventh million new ones. Truly, the “last 60 days” count seems pegged at 1.7 million and change—will it change sometime soon? There has been plenty of progress on the technical side; I’ve really enjoyed a couple of voice-enabled visits with folks meeting around Dr. Dobb’s Journal venues. Getting voice to work was not trivial at first, but neither was it a real problem once turned on to Control-P.

I’m starting to try and get the Open Metaverse stuff going. Looks like I’ll be trying to have Visual Studio 2005 Express compile a client soon.

Nothing like a user conference to motivate poster production! Somehow the chance to share work with perhaps 20,000+ eyeballs at the San Diego Convention Center always adds a bit to the excitement. There are now eight 3-foot by 4-foot color posters that emphasize shots of the progress made on the land surface, buildings, street signs, street lighting, sidewalk lighting, and foliage.

Oh, and there’s one other panel that has SL snapshot images for decoration, but is really a little manifesto of the importance of metaverses (in 2007) to the future of spatial systems.

This is Darb’s manifesto posted for attendees in the Map Gallery of the ESRI International User Conference in the Sail Room of the San Diego Convention center, 18–22 June 2007
The attendees are geographic information systems professionals, managers, and supporting industry folk who largely work with maps, map servers, and related technology for a living.

—————————-<>—————————-
YOU WILL SOON WANT A METAVERSE FOR YOUR SPATIAL DATA

Metaverses are immersive 3D computer graphics platforms
- They are not too much like 2-1/2D raised terrain or globes.
- Their objects may not support the vertex model of GIS or CAD,
but use parametric points or U-V maps and raster textures instead
- Through a viewer or other tools, metaverses immerse the user into the 3D model.
- An immersed user is as likely to look up or under as a globe user is to look downward.

If the metaverse holds a model built honoring GIS data, then a metaverse might
- Place the user into the map
- Allow one to stroll through a geodatabase
- Publish spatial data in real-time 3D for very many simultaneous users

Metaverses can allow massively simultaneous at-will rendering in near-real time
- As an example, Second Life is built on grid computing with >5000 processor cores
- Second life spatial data are integrated parametric point objects and raster textures
(34 Terabytes as of 5 May 2007)
- Second life supports over 40,000 simultaneous users worldwide with streaming audio and video.
Integrated VOIP is in beta.

Open-source options exist for single regions, and are developing for grids
- Second Life’s producer, Linden Lab, has announced plans to open source their server code
- This would allow cost of hardware / server power / model development to become the
limiting factors for a civic-scale metaverse
- City of Berkeley could stand up a 1:1 scale immersive model on about 512 processor cores,
or 1k cores with a redundant grid

Metaverses typically include a physics engine
- this manages object collisions and optionally provides gravity and
- in Second Life, the physics engine in each processor core handles collisions among
up to 15,000 objects in the core’s region.
- the engine does so at 40 Hz (forty cycles per second) to allow rendering throughout
the region as real-time movies for each client.

Metaverses will change your data center expectations
- There will be a desire to build out grid computing
- Performance will be tied to processor cores, while most related resources such as
system memory and disk storage (per core) are not exceptional
- In metavserses, the simulated space expands linearly with the number of regions in your grid.
Second life has 64K square meters, about 16 acres, at 1:1 scale, for each processor core
- people interested in grid computing are very interested in having processors with more cores
- these people may be equally uninterested in having operating system costs, or even server application costs, scale with the number of cores

SL Darb Dabney, Berkeley, California 20070615