PACET stands for PC-Parallel Architecture for Cost Efficient Telemetry Processing. The objective of this project is to investigate an efficient high-performance solution for telemetry processing which requires high-performance data processing for the telemetry downlink. The downlink data rates is expected to reach 50 Mbps and beyond with the new Earth Observing System (EOS) era of satellites (see Earth Observing System (EOS) Data and Information for more details). Our predecessor project is based on VLSI/LZP and the fine-grained parallelism approach implemented on the MasPar MPP (MP1 and MP2).
The PACET project employs a distributed computing approach. The PACET system is a loosely-coupled parallel computer consisting of high performance Unix-clone workstations connected through a high speed FastEthernet network. The implementation is projected to achieve very high performance at a low cost by integrating the power of multiple COTS (Commodity Off The Shelf) PC clones orchestrated to operate in parallel similar to the project Beowulf. Due to the difference in the architecture, coarse-grained parallelism must be utilized.
PACET will perform identical functions to its predecessor. However, since PACET architecture is different from the MasPar architecture, new design and implementation must be considered. The following algorithms illustrate the MPTP SIMD-based algorithm and the PACET PVM-based algorithm. Although the number of computing nodes in PACET is much smaller than the number of PEs in the MasPar, each PACET's computing node has a much higher performance and more memory capacity.
