摘要或動機

The research is about the optimal on parallel processing. Through boot disk – which will automatically finish booting configuration, .it is efficient and quick to build high performance PC clusters. The advantage of parallel computing could be applied to massive image processing. By sharing processing and breaking huge processing load into lots of pieces, we could get more efficient result. It is also possible to optimal parallel system through some special means such as dynamic configuration. Through the means, the system could distribute work loading itself. It could also adjust itself to get the highest performance and the most stable environment. 本研究之目的在於探討平行處理中的計算資源的最佳化，透過自動完成開機設定的Boot Disk 來有效快速建製出高效率的PC Clusters 環境，並透過動態配置與類神經網路的校調，使整體叢集的運算能自動調整至最佳化。平行處理優勢，可以應用在耗費極大量的運算資源的影像處理上。透過運算資源分享，可以以很高的效率將極為龐大的運算工作分散成許多較小的程序，使影像處理速度加快。經由平行演算法及實際應用的調整，可對已成形之平行系統作效能上的加強。使用類神經網路的方式訓練，使其系統能夠自我分配運算工作量，且隨著各平行化程式與各節點的不同，能自我校調至最佳化，達到高效率且穩定的運算環境。\r
本研究透過高效率且能自我調校的運算環境，可用於優化其本身結構，以達到\r
演化出更進一步系統，具有相當大的發展潛力。