隨著雲端計算的蓬勃發展，伺服器整合允許個人用戶或是公司充
分利用他們現有計算資源並且消除資源過度配置的發生。在這種狀況
下記憶體的使用會是一個非常關鍵的因素：它不只決定了宿主機能夠
容納下多少台虛擬機器，同時也會影響到客戶機和宿主機的效能。
分頁共享是一個常見的內存過量使用技術：宿主機上重複的分頁
會被偵測到然後會共用一個寫入保護的分頁，如此那些共享中的分頁
記憶體的就可以被回收使用。KSM(Kernel Samepage Merging)是一個
從 Linux 2.6.32 以來內建的去冗餘程式。本論文提出了 pre-HZ
KSM(pre-identified Hotzone KSM)。記憶體熱區是在記憶體上的一段空
間，存在其中的分頁有非常高的機率能夠被合併和共享。我們旨在利
用記憶體熱區的特性來優化 KSM 的效率。我們藉由熱區表來判斷出熱
區分頁，之後修改 KSM 的掃瞄機制使其能夠優先掃描這些熱區分頁。
實驗證實 pre-HZ KSM 相較於原本的 KSM 能夠節省約 20%至 30%的時
間，並且不會耗損太多額外的 CPU 資源。

Server consolidation allows users to better utilize their computing resources. The most critical factor of consolidation is memory provisioning because it does not only determine how many virtual machines (VMs) can be accommodated on the host physical machine, but also affects the performance of guests and host. Same-Page Merging is a common technique of memory overcommitment, which detects duplicated pages and shares single write-protected copy of page among VMs. KSM (Kernel Samepage Merging) is a built-in memory de-duplication program in Linux since 2.6.32. In this thesis, we proposed pre-HZ KSM (pre-identified Hotzone KSM), whose goal is to optimize the performance of KSM by exploiting memory hotzones. A memory hotzon is a set of pages that have higher chance to be merged. We identified the hotzone pages by looking up the pre-build hotzone table and modify the scanning pattern to scan those hotzone pages prior to the non-hotzone pages. Experiments show that pre-HZ KSM can accelerate the merging speed 20\% to 30\% faster than vanilla KSM without consuming too much additional CPU resource.

Chinese Abstract............i
Abstract...................ii
Contents..................iii
List of Figures.............v
List of Tables.............vi
List of Algorithms........vii
1 Introduction..............1
2 Background................3
3 Pre-HZ KSM...............10
4 Experiments..............20
5 Conclusion...............28

[1] Andrea Arcangeli, Izik Eidus, and Chris Wright. Increasing memory density by
using ksm. In In OLS, 2009.
[2] Sean Barker, Timothy Wood, Prashant Shenoy, and Ramesh Sitaraman. An
empirical study of memory sharing in virtual machines. In Presented as part
of the 2012 USENIX Annual Technical Conference (USENIX ATC 12), pages
273–284, Boston, MA, 2012. USENIX.
[3] Edouard Bugnion, Scott Devine, Kinshuk Govil, and Mendel Rosenblum. Disco:
Running commodity operating systems on scalable multiprocessors. ACM
Trans. Comput. Syst., 15(4):412–447, November 1997.
[4] C. R. Chang, J. J. Wu, and P. Liu. An empirical study on memory sharing
of virtual machines for server consolidation. In 2011 IEEE Ninth International
Symposium on Parallel and Distributed Processing with Applications,
pages 244–249, May 2011.
[5] Licheng Chen, Zhipeng Wei, Zehan Cui, Mingyu Chen, Haiyang Pan, and Yungang
Bao. Cmd: classification-based memory deduplication through page access
characteristics. In ACM SIGPLAN Notices, volume 49, pages 65–76. ACM,
2014.
[6] Yan Deng, Chunming Hu, Tianyu Wo, Bo Li, and Lei Cui. A memory deduplication
approach based on group in virtualized environments. In Service Oriented
System Engineering (SOSE), 2013 IEEE 7th International Symposium
on, pages 367–372. IEEE, 2013.
[7] Djellel Eddine Difallah, Andrew Pavlo, Carlo Curino, and Philippe Cudre
Mauroux. Oltp-bench: An extensible testbed for benchmarking relational
databases. Proc. VLDB Endow., 7(4):277–288, December 2013.
[8] Konrad Miller, Fabian Franz, Marc Rittinghaus, Marius Hillenbrand, and Frank
Bellosa. Xlh: More effective memory deduplication scanners through cross-layer
hints. In USENIX Annual Technical Conference, pages 279–290, 2013.
[9] Yeji Nam, Dongwoo Lee, and Young Ik Eom. Self: Improving the memorysharing
opportunity using virtual-machine self-hints in virtualized systems. In
Proceedings of the 6th Asia-Pacific Workshop on Systems, page 10. ACM, 2015.
[10] Te-Yu Tsai. Improving start-up performance in kernel samepage merging based
on memory hot zone. unpublished, 2017.
[11] T Veni and S Bhanu. Mdedup++: Exploiting temporal and spatial pagesharing
behaviors for memory deduplication enhancement. The Computer Journal,
59(3):353–370, 2016.
[12] Carl A. Waldspurger. Memory resource management in vmware esx server.
SIGOPS Oper. Syst. Rev., 36(SI):181–194, December 2002.
[13] Nai Xia, Chen Tian, Yan Luo, Hang Liu, and Xiaoliang Wang. UKSM: Swift
memory deduplication via hierarchical and adaptive memory region distilling.
In 16th USENIX Conference on File and Storage Technologies (FAST 18), pages
325–340, Oakland, CA, 2018. USENIX Association.