|Table of Contents|

Media File Replication Strategy Based on Access Statistics


Research Field:
Publishing date:


Media File Replication Strategy Based on Access Statistics
SONG Xiao-hua1HUANG He-qing2CAO Yuan-da1
1.School of Computer Science and Technology,Beijing Institute of Technology,Beijing 100081,China;2.North China Power Dispatching Transaction Center,Beijing 100053,China
clusters Zipf d istribut ion file replicat ion replica creat ion load ba lance
This paper uses statistica lmethod to pred ict the popu larity of the files based on the access statistics and decides on the number o f the replicas according to the popu larity and the attributes of the file. Tak ing the attributes o f different servers into accoun,t the rep lica is placed on the co rresponding servers by themeans of dynam ic intervalmapping. The exper imenta l results show that th is replication strategy improves the performance and assures the load balance.


[ 1] Jonathan D, Jerem y G. Dynam ic repacking: a content replication policy fo r c lustered m ultim edia servers [ R ]. Dublin: T rin ity Co llege Dublin Computer Sc-i ence Departm ent Techn ica l Reports TCD - CS - 2002 - 36, 2002.
[ 2] X iao Bozhou, Cheng Zhongxu. Optim a l v ideo rep lication and p lacement on a cluster o f v ideo on dem and servers [ A ]. Proceed ing s o f the 2002 Inte rnational Conference on Pa ra lle l Processing [ C ]. London: The 2002 Inte rnational Conference on Paralle l Process ing, 2002. 238- 247.
[ 3] A lex F B. Fundam enta ls o f theM onte Carlom ethod for neutra l and charged particle transpo rt [M ]. M ich igan: The University ofM ich igan Press, 2001.
[ 4] M atsum otoM, N ishim ura T. M ersenne Tw ister: A 623- dim ensionally equ-i distributed un iform pseudorandom number genera to r [ A ]. Proceed ings of the ACM Trans onM ode ling and Computer [ C]. Syndney: ACM Trans onM ode ling and Computer, 1998. 195- 203.
[ 5] http: / /www. powe rec. net/m arke tp lace / shuzitushuguan/ bzsjk. html .


Last Update: 2007-10-30