Parallel servers visualize satellite data
-Andrew Wilson
RI International (Menlo Park, CA) is using a distributed-parallel-storage-server (DPSS) architecture that was developed at Lawrence Berkeley National Laboratory (Berkeley, CA) to explore digital landscapes. In operation, the TerraVision system can display landscapes created from elevation data and aerial images. Data for visualization resides on DPSSs distributed across an ATM network.
According to Brian Tierney, one of the codevelopers of the DPSS, the distributed nature of the DPSSs and the algorithms used by TerraVision to prefetch and display data allow users to roam about a terrain composed of hundreds of gigabytes of data. The DPSS is composed of disk servers that operate in parallel over a wide-area network to provide access to high-speed, large data sets. To achieve high performance, multiple low-cost, medium-speed disk servers use the network to aggregate data streams. Data blocks are dispersed so that many system elements can operate simultaneously. This allows a large collection of disks to seek in parallel and all system servers to send data to the application in parallel.
A typical DPSS server consists of a UNIX workstation with four to six SCSI disks and a high-speed network interface that is capable of operation at 60 to 120 Mbits/s. "Using several DPSS servers, a data pipe can be constructed that is fast enough to satisfy almost any application, " says Tierney.
For more information on distributed parallel storage servers see http:// www-itg.lbl.gov/DPSS or contact Brian Tierney at BLTier [email protected].
