An Improved Method of Cache Prefetching for Small Files in Ceph System

  • Ya Fan
  • Yong Wang
  • Miao YeEmail author
  • Xiaoxia Lu
  • YiMing Huan
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 891)


To improve the inefficiency of file access for massive small files in the distributed file system, prefetching the small files into the cache is the most conventional method to be adopted. As the number of cached files is proportional to time, there will be many redundant small files which occupy the cache space and haven’t been read for a long time, the hit ratio will be decreased in this situation. To solve this defect, we proposed an improved LRU-W algorithm based on the file read times and the time interval of file read and designed a L2 cache optimization mechanism which can search file in the linked list with higher priority firstly and remove the files with lighter weight factor from the linked list with lower priority dynamically. The experiments and its result show that when prefetching massive small files, the proposed method in this paper can increase the hit ratio of cached files and improve the overall performance of Ceph file system.


Massive small files Ceph LRU-W L2 cache 



This work is partly supported by the National Natural Science Foundation of China (Nos. 61662018, 61861013), the Project of Science and Technology of Guangxi (No. 1598019-2), Foundation of Guilin University of Technology (No. GUTQDJJ20172000019).


  1. 1.
    Beckmann, N., Sanchez, D.: Modeling cache performance beyond LRU. In: IEEE International Symposium on High PERFORMANCE Computer Architecture, pp. 225–236. IEEE Computer Society (2016)Google Scholar
  2. 2.
    Jaleel, A., Najafabadi, H.H., Subramaniam, S., et al.: CRUISE: cache replacement and utility-aware scheduling. ACM SIGARCH Comput. Arch. News 40(1), 249–260 (2012)CrossRefGoogle Scholar
  3. 3.
    Jung, D.Y., Lee, Y.S.: Cache replacement policy based on dynamic counter method. Adv. Sci. Lett. 19(5), 1530–1534 (2012)CrossRefGoogle Scholar
  4. 4.
    Jiang, B., Nain, P., Towsley, D.: LRU cache under stationary requests. In: ACM Sigmetrics Performance Evaluation Review, vol. 45(2) (2017)CrossRefGoogle Scholar
  5. 5.
    Perkowitz, S.: A survey of Web cache replacement strategies. ACM Comput. Surv. 35(4), 374–398 (2003)CrossRefGoogle Scholar
  6. 6.
    Ding, J., Wang, Y., Wang, S., et al.: Design and implementation of high efficiency acquisition mechanism for broadcast audio material. In: IEEE/ACIS, International Conference on Computer and Information Science, pp. 667–670. IEEE (2017)Google Scholar
  7. 7.
    Niu, D.J., Cai, T., Zhan, Y.Z., et al.: Metadata caching subsystem for cloud storage. Appl. Mech. Mater. 214, 584–590 (2012)CrossRefGoogle Scholar
  8. 8.
    Huang, X.Y., Zhong, Y.Q.: Web cache replacement algorithm based on multi-markov chains prediction model. Microelectron. Comput. 5, 123–125 (2014)Google Scholar
  9. 9.
    Wang, W.: Research on Web Cache and Prefetching Model Based on Access Path Mining. Southwest Jiaotong University (2014)Google Scholar
  10. 10.
    García, R., Verdú, E., Regueras, L.M., et al.: A neural network based intelligent system for tile prefetching in web map services. Expert. Syst. Appl. Int. J. 40(10), 4096–4105 (2013)CrossRefGoogle Scholar
  11. 11.
    Jing, C., Wang, M., An, P.C., et al.: 3D model prefetching system based on neural network. Comput. Appl. Softw. 32(7), 182–185 (2015)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Ya Fan
    • 1
  • Yong Wang
    • 1
  • Miao Ye
    • 2
    • 3
    Email author
  • Xiaoxia Lu
    • 2
  • YiMing Huan
    • 2
  1. 1.School of Computer Science and Information SecurityGuilin University of Electronic TechnologyGuilinChina
  2. 2.School of Information and CommunicationGuilin University of Electronic TechnologyGuilinChina
  3. 3.Guangxi Colleges and Universities Key Laboratory of Cloud Computing and Complex SystemsGuilin University of Electronic TechnologyGuilinChina

Personalised recommendations