隨著計算資源成本的降低和網際網路的普及，建置分散式系統的門檻已大大的降低了，許多以分散式計算為基礎的新型服務也應運而生，因此設計分散式系統時對於自動化管理和可程式化的需求也因此而提高了；為了提高實作分散式系統時的便利性我們引入了主動物件計算模型的概念，希望能藉此降低實作分散式程式的困難度。此外，我們也建議使用自動化記憶體管理機制來減輕人為控制可能產生的風險並進而提昇系統的穩定度。 然而，主動物件計算模型的特性可能會使垃圾收集機制發生誤差，例如，不同步訊息傳遞會使垃圾收集演算法產生競賽情況，這會降低自動化記憶體管理的正確性。因此為了使垃圾收集機制能完全整合在主動物件計算模型之中，我們提出了權重參照計數演算法，並針對可能的錯誤做了必要的修改。最後，我們會藉由比較使用各種垃圾收集機制時程式的執行速度和可用記憶體的容量來彰顯權重參照計數演算法的成效。 With the lower of the cost of computation resource and Internet becomes more popular, the idea of distributed computing is widely used in several applications. As the result, automatic management and programmability are become more desirable while implementing a distributed system. Actor Model has several practical features, and we believe these features would be helpful when designing a distributed system. Moreover, automatic memory management can be used to handle the problem of memory leakage caused by erroneous memory operation, so we suggest to apply automatic memory management to Actor Model. Here we use the term Garbage Collection to present the concept of automatic memory management. However, there are some features of Actor Model are harmful to the correctness of Garbage Collection algorithm. For example, asynchronous message passing would causes race condition, then the Garbage Collection mechanism is flawed. So we propose Weighted Reference Counting as our Garbage Collection algorithm, and add a necessary mechanism to improve fault tolerance. At last, we implement Weighted Reference Counting on SALSA programming language.
