Munenori Kai

The authors already revealed that a user-level remote memory paging system, which provides a larger size of memory beyond that of local physical memory by using remote memory distributed over cluster nodes, is higher performance and more stable than a kernel-level one. It requires an efficient page replacement algorithm because the remote page swap is critical in performance for using the remote memory. However, an user-level software can be hardly apply its page replacement algorithm using memory access information in OS. If it uses pseudo memory access information, it can be applied the efficient page replacement algorithm. But this cost and performance are not evaluated sufficiently. In this paper, we implement and evaluate pseudo memory access bits, and reveal that it can be used actually. Furthermore, we compare page replacement algorithms using them and not using, the Clock algorithm, the NRU and pseudo-LRU, Simple algorithm, page replacement algorithm using swap-in history, the Random algorithm and the FIFO.

The authors already proposed the Distributed Large Memory System: DLM which provides a larger size of memory beyond that of local physical memory by using remote memory distributed over cluster nodes. The original DLM adopted the Clock algorithm as a low cost page replacement algorithm. In the DLM, the remote page swapping is the most critical in performance. For more efficient swap-out page selection, we propose the new page replacement algorithm which pays attention to swap-in history: CASIN (Clock Algorithm with Swap-IN history). The LRU and other algorithms which use the memory access history generate more overhead for user-level software to record memory accesses. On the other hand, using swap-in history generates little costs. According to our performance evaluation, this new algorithm has reduced the number of the remote swapping in the maximum by 32% and gains 2.7 times higher performance in real application, Cluster3.0. In this paper, we describe the mechanism of the new page replacement algorithm and show a performance evaluation for three applications.

Authors designed and evaluated Distributed Large Memory System: DLM, which gives us very large virtual memory beyond the size of local memory by using remote memory distributed over cluster nodes. Originally, the DLM only supports a function for dynamic memory allocation, but it is required to be equipped with a memory release function for the applications that reuse memory resources in their runtime. We incorporate a simple memory management mechanism into the DLM to support memory release function for efficient memory utilization. In this report, we describe the design and mechanism of the memory management and show an initial performance evaluation for a red-black tree as a micro benchmark and a cluster 3, a real bioinformatics application.

Programming lessons in universities start in a language widely used in practice, such as Java, of which richness in syntax and semantics tends to be rather an obstacle for students to learn and enjoy programming. We developed a programming language Nigari and its supporting system to present students an environment where they learn OOP fundamentals by writing attractive examples to be visualized on the screen from the beginning. Nigari is designed to lead to writing programs in Java. We report on our class experiment of applying Nigari and its system at the CS department in Waseda University.

大学のプログラミングの授業では，Javaなどの既存のプログラミング言語を題材とすることが多い．それらは習得が難しく，プログラミング嫌いを生む原因となる．初学者には，興味を惹く例題がすぐに書け，それでいて既存言語へ容易に移行できるプログラミング言語と環境を用いるのが適切である．そこで，オブジェクト指向のプログラミングにすぐ入れて，なおJavaへの移行も容易な%このような性質を持った言語Nigariとそのプログラミング環境を実際に開発して，授業に用いてみた．本発表では早稲田大学コンピュータ・ネットワーク工学科での，その試行結果を報告する．Programming lessons in universities start in a language widely used in practice, such as Java, of which richness in syntax and semantics tends to be rather an obstacle for students to learn and enjoy programming. We developed a programming language Nigari and its supporting system to present students an environment where they learn OOP fundamentals by writing attractive examples to be visualized on the screen from the beginning. Nigari is designed to lead to writing programs in Java. We report on our class experiment of applying Nigari and its system at the CS department in Waseda University.