SDMG Notes for 16 March 2006

In attendance:

Theresa V, Linda A, Tim F, John W, Keith O, Jerry M, Terralyn V, Kelly C, Matt G, Dave H-W (notes)

Special Topic: How to separate processes to run on different processors (mini cluster computing)

Terralyn discussed the existing cluster computing system at the FSL, which is presently not used to capacity. The cluster consists of 32 separate PCs (called “nodes”) that are running Linux. One is the master, or controlling, node, and the rest are “compute” nodes. The cluster runs a queuing system called the Parallel Batch System (PBS) . To use the cluster, jobs must be separated by the user into distinct processes, which then produce separate output files. PBS farms each one out to different nodes, and then the user needs to stitch the output files back together. In this system, nodes cannot communicate with one another; a node works on one spatial location and is oblivious to the processes running in all neighboring locations. ArcInfo cannot run on this system, but GDAL, which can use multiple spatial formats, including ESRI, can (ask Matt if you’re interested).

In contrast, the College of Engineering has a high-powered cluster ( with individual nodes that have 4-10 GB RAM each ) that has the ability to utilize “shared memory,” which means that one large problem can be loaded, and all nodes will be utilized and can communicate with each other. Their cluster is somewhat underutilized, so COE is open to other research groups joining their system, as long as newcomers buy new nodes to help expand the cluster’s capacity.

If people need to use a “shared memory” cluster, joining COE’s cluster is probably the way to go. The College of Forestry is unlikely to have such a cluster in the near future. Managing multiple clusters could be difficult and unnecessary, so Terralyn will look into the possibilities of setting up multiple queues or entry points into the cluster, so that more than one group can access the cluster at one time. So far we haven’t had enough users to warrant this, but there are enough nodes in the cluster to handle multiple queues.

Resources from the web:

  • Spatial Technologies - Clustering Computing and Storage
  • Decentralized GIS Web Services on Grid

    Treats: Terralyn
    Notes: Dave

    Meetings:

  • Scripting meeting in PDX (report):
    Theresa attended. They’re planning to meet every three months or so.
  • Python Group:
    The installation notes for the latest version are available. No set date for the next meeting.
  • FS Remote Sensing Application Conference: April 24-28 in Salu Lake City,UT
  • GIS In Action: April 18th and 19th in Vancouver, Washington
  • ESRI User Conference in San Diego August 7-11, 2006
  • Where 2.0 Conference: San Jose, CA June 13-14: The Where 2.0 Conference brings together the people, projects, and issues leading the charge into the location based technological frontier. url

    Training:

  • Spatial Stats at Seed Production Research Center: happening at meeting time
  • No other needs identified

    2005 Digital Ortho Quads:

    One-meter color DOQs (county mosaics) have been produced. They have not been edgematched. The state is working with the Institute for Natural Resources to figure out how to distribute these to the public. It will probably be at least September before they are available online. They will eventually have DOQs from multiple years available for comparison purposes. The photos were taken with film (not digital). The state will produce ½-meter color DOQs from the same film after the 1-meter versions are completed.

    Link to site to download county mosaics: NRCS Data Gateway
    Theresa has Benton and Lane Counties available on portable disk drive.

    Round Table:

  • Matt: His group is hiring an FRA to assist with the three-state GNN project. They are interviewing five candidates during the week of March 20. He also recommended SWIG, a “geeky programming tool” that compiles C++ scripts, allowing them to be used in other scripting programs.
  • Keith – Getting photorealistic visualizations of forested landscapes going. He is using Visual Nature Studio 2 which is an upgrade to World Construction Set. The program is expensive ($2000) and difficult to learn, but produces nice results. The software is made by 3D Nature.
  • Dave: Renewing his work in the TELSA/VDDT modeling world (more next month).
  • Keith: Likes the brighter plotter paper. Freedom Paper: Source for plotter paper.
  • Tim: Planning to go to the ESRI Developer Conference and O’Reilly Where 2.0 (Location Based Services).
  • John: Mentioned the GRB data format, the only format that WMO (World Mapping Organization?) is producing.
  • Linda: She is becoming the Hinkle Creek data manager, will be doing field work on the Trask project (Slime Queen), and recommended that we check out John Bolte’s Evoland, which is up and running for the McKenzie River basin.
  • Theresa: She has been Metadata Queen lately.

    Next Meeting: April 20th
    Speakers:

  • Dave to give his talk "Spatially explicit modeling of alternative futures for forest landscapes using TELSA/VDDT "

    Notes: Matt
    Treats: John