本研究提供了一個改善郵件過濾效能的方法,並以實驗模擬的方法比較每一進程套用最佳的數量s後,與原來的每一過濾進程數量s=1的整體過濾效能差異。因每一個過濾進程使用同一硬體運算資源,共用CPU資源會產生背景切換(Context Switch)的影響,因此整體過濾效能提升幅度無法達到理想值。而s總數的上升過多時,因過於頻繁的CPU背景切換,整體過濾效能上升也會受限,故本研究再藉由設定每個過濾進程s數量的上限分別進行實測,以求取一個可讓整體過濾效能提高,但較少CPU背景切換影響的s數量配置。 The spam spreading problem was getting worse continuously. Many people and enterprises took the antispam mechanism as their solution, but antispam mechanism relies on multiple filtering procedures to improve the accurate rate of determination. This may caused delayed delivery. This research tried to improve each filtering procedure performance, and raised whole filtering performance to reduce the delay of e-mails. According to the actual average e-mail arrival rate of λ, we fixed the minimal waiting line among filtering procedures and calculated the better service numbers of procedures "s" based on the formula of Queueing Theory.
We provided a method of improving filtering performance and compared filtering performance difference between processing theoretical optimal service numbers of "s" and the original "s=1". Each procedure shared the computing resource of the same hardware, context switch may result in the improving of whole filtering performance could not match the theoretical optimal value. When numbers of procedures "s" increased, this will cause frequently context switch of CPU and limit the improving of filtering performance. Therefore we perform the experiment by restricting the numbers of "s" trying to find which number of service processes could improve the whole performance but less context switch.