Saturday, March 1, 2008

Installation of Quadscreen Virtual Reality Workstation at the Burnham Institute



This sign welcomes Burnham02 Alter to SL. It is strategically placed at the home base of the virtual workstation, in my lab in Second Nature.

Ramses took this one...after 3 days of work, the workstation was no longer stalling.


Jeff Price and Ramses, discussing Second Life, with Burnham02 Alter hovering in front of my ACS-SL CAIA lab. The virtual reality workstation itself is strategically placed right next to the fridge.

This past week I went to the Burnham Institute in La Jolla, CA, to install a quadscreen Virtual Reality workstation powered by SL, in the lab of Dr. Jeff Price. This was an interesting visit, as the Burnham Institute is the site for one of the Molecular Libraries Screening Network Centers, part of the National Institutes of Health initiative in Molecular Libraries. Jeff Price is a prominent figure in high content screening instrument development. He was the founder and CEO of Q3DM, one of the original high content screening companies, which was acquired by Becton Dickinson several years back before Jeff moved on to academia and then to the Burnham. At the Burnham, Jeff is in charge of the high content screens of the MLSCN center. When I was at Cellomics, Jeff and the other scientists at Q3DM were one of our competitors....so this was an interesting visit, for sure.



The installation of the Virtual Reality workstation came about as part of the collaboration between MACE -The Michigan Alliance for Cheminformatic Exploration- and the MLSCN center at the Burnham. Marc Mercola, Jeff Price and the other scientists at the Burnham have always been very supportive of the interaction between the Chemiformatic Centers and the Screening Centers so they welcomed me to their Center and they were more than happy to see what it was that I was going to show them. Well, as soon as I showed them the ACS CAIA, Jeff and the other members of his high content screening group were immediately impressed by the powerful graphics of SL, as well as the ability to interface with the wikis so that images can be annotated by human users as well as by machines. Jeff, and everyone else in Jeff's lab quickly grasped the importance of this technology in terms of image data visualization and communication. Fortunately, itt really does not take too much effort to convince the experts of how good and powerful SL graphics capabilities are... so the visit can be declared a success!

As expected, the installation of the workstation was not a simple matter, involving a lot of blood, sweat and tears. Immediately after I installed it and began testing it, I found out the workstation was freezing when the avatar went into the ACS CAIA. I was able to figure out the problem after two days of troubleshooting, and then came up with a work around: I found out that the memory usage of SL was maxing out at about 400MBytes, wereas in my laptop SL memory usage would not max out until reching about 700MBytes, within the same sim. The maxing out would cause the quadscreen to stall, whereas my laptop would cruise through the sim without any glitches because its memory was not maxing out. I found out there was plenty of RAM memory left on the quadscreen, so the maxing out of the memory was not due to the lack of RAM (I put 4G Bytes RAM on the quadscreen same as my laptop, and I was only using about a fourth of it, with 1.2 GBytes left).

Looking further into the hardware components, I found out that the NVIDIA graphics cards that went into the workstation had 256 MBytes memory each, while the graphics card powering my laptop had 512 MBytes memory. I believe this is the root cause of the problem. As a workaround, I limited the memory cache in the SL setting to 64MBytes. This, plus other adjustments on the SL graphics settings (such as setting the maximum sight distance to less than 120 meters) improved the quadscreen performance so that one can navigate around the ACS CAIA without stalling or freezing. Nevertheless, there is some lag everytime the avatar moves to a new view on the CAIA, since the images have to be loaded through the network as opposed to being stored in the graphics card memory. But, this lag is only apparent if one has been cruising around the ACS CAIA from a better computer...like my laptop. So, I figured I wouuld leave the virtual reality workstation as is. If anyone at the Burnham is actually going to put it to good use, I will work to upgrade the workstation's graphics cards, so that they get best experience possible. For the time being, I will work on getting the Burnham scientists set up with some avatars, help them through "disorientation island", and then set my SL lab as home so I can tour them around and guide them through the world when they decide to enter again on their own.

4 comments:

Peter Miller said...

Do you upload multiple single images or composites that you then offset to display the requisite part thereof? I've no personal experience but understand that the latter is normally more efficient and, of course, cheaper. It might, however, require some preprocessing prior to upload but uploads would be cheaper and faster.

Gus Rosania said...

Peter,
Excellent points. The images were uploaded individually, using the bulk upload function within SL. Actually, images in a CAIA are ultimately meant to be moved around and sorted based on the different chemical or visual features, so I think that uploading the images as a composite file would not be a good idea. On the other hand, because one has to pay for the individual uploads, I do not think that one would want to upload millions of images without actually intending to visualize them. So the current upload pricing schemes works if one only uploads those images one is really interested looking at. Beyond that, if one wants to upload millions of images up front without considering if they are ever going to be viewed, then someone is going to have to negotiate a prize with Linden Labs. A different possibility would be use a different software package that would allow visualizing images stored in ones own server, without having to pay an image upload fee. Any ideas?

Peter Miller said...

I have no idea how many images are on the LL servers but uploading further "millions" is something you should almost certainly discuss with them. Your application is evidently serious and valuable and it would be unfortunate if the system subsequently proved unable to support it adequately. The pricing on uploads, as I understand it, is not to generate a profit so much as to deter frivolous uploading that would place a load on the asset server (as with many aspects of SL, the system is hobbled slightly to optimise the experience for the average user and make it harder to grief). I have relatively little experience of the new VWs coming along but from what little I've seen would suggest trying the realXtend implementation of OpenSim which is easy enough to install though my laptop is underpowered to run both server and client simultaneously. The "feel" of the system is very close to SL. You might want to discuss implementation of your current SL system with someone like Eloise Pasteur (who I believe did the terraforming for ACS and is well-known to Horace). She would be much better able to advise on strategies.

Gus Rosania said...

Peter,
Thank you for your advice. Making microscope image data from high throughput screens available for everyone to analyze and explore is a big challenge. I think SL (or some other metaverse application) can solve the front-end visualization and communication problem. Nevertheless, there are other problems in terms of the best way to interface with the instruments acquiring the data and the machine vision, cheminformatic and statistical analysis tools that help make sense of the data. That is why setting up the ACS CAIA and getting the Virtual Reality workstation set up here the Burnham is important, but it is just a first and second step...I am hoping the Burnham's scientists will see this as an opportunity to (potentially) solve both front and back end problems. With their support, I may be able to get some additional leverage from NIH, NSF or some other agency.

About Me

I am Assistant Professor at the University of Michigan College of Pharmacy, Department of Pharmaceutical Sciences