Research on Improved Cache Replacement Algorithm Serving for Wind Power
System
Erjie Lu, Luan Chen, Jian Li, Bintao Mo, Jian Chang, Qi Huang
School of Energy Science and Engineering, University of Electronic Science and Technology of China
Chengdu, Sichuan Province, China
e-mail: 549008743@qq.com; chluan@uestc.edu.cn; leejian@uestc.edu.cn; 103478201@qq.com;
970781286@qq.com; hwong@uestc.edu.cn
Abstract—To improve the distributed cache replacement
algorithm performance in wind power system, after
researching several traditional algorithms such as LFU (Least
Frequently Used), SIZE (Object Size Based), LRU (Least
Recently Used) and Hybrid (Saved Value Based), a new
algorithm FST (Frequency, Object size, Access time) is
proposed. This algorithm combines the features of access
frequency, object size, access time interval, and the longest
time without access. It avoids the outdated “hot spots”
document’s occupying memory space to waste resources, so it
is supposed to improve the system performance. By taking
contrast test with LFU, SIZE and LRU in the wind power
system cache server, FST algorithm shows better performance
in request hit rate, byte hit rate and reducing delay time.
Keywords-cache replacement algorithm; wind power system;
request hit rate; byte hit rate; delay time
I. INTRODUCTION
As the rise of the turbine capacity, the proportion of wind
power in new energy power grid increases gradually, so the
analysis of wind farm data is very important to grid operation.
A wind farm can produce millions of data one day. Then
how to manage and store the huge amounts of data
efficiently is a problem to be solved. Reference [1] designs a
large wind farm database management system based on SQL
Server. Based on the design, reference [2] optimizes the
querying statements of the traditional database to improve
the response speed. Although the above scheme has an
improvement in a certain extent, with the grid system
becoming more complex and the amount of data increasing
rapidly, centralized storage tend to show some problems
such as database overflow, low reliability and scalability. A
method of applying
distributed storage in the research on
massive data is proposed in reference [3]. Taking advantage
of the parallel operation idea of MapReduce and Hadoop’s
characteristics of high reliability, large data processing
capacity, flexible expansion and low cost, the problems of
traditional database like insufficient capacity can be solved
effectively.
In wind power system’s database server, if we get the
access request from the original database every time, it will
not only increase the user’s access delay time, but also add
the load of the server and cause network congestion
problems [4]. Therefore, storing the data which has been
accessed in the computer’s local storage space by the cache
technology will effectively improve the overall performance
of the system. When the data is accessed again, it can be
gotten from cache directly instead of requesting to database
server repeatedly. However, the capacity of each cache
system is limited, when the amount of data in the cache
exceeds the system’s memory, it is appropriate to replace
some of the data in the cache to ensure the new data can be
added [5].
Cache replacement algorithm is the key factor that affects
the performance of the cache. A good replacement algorithm
can guarantee the data in the cache a high hit rate. Reference
[6] introduces several traditional cache replacement
strategies and their respective characteristics. The most
common replacement algorithms are LFU, SIZE, LRU and
Hybrid. LFU is based on the access times of object in cache.
SIZE is based on the object size. LRU is designed based on
temporal locality principle. Hybrid the most typical
representative of the replacement algorithms based on saved
value. Reference [7] proposes several important indexes for
evaluating cache replacement algorithms. Reference [8]
presents an improved method based on the Hybrid cache
algorithm, but it is too complicated to realize. Based on these
findings, this paper designs a cache replacement algorithm
FST integrating access frequency, object size, access time
interval and the longest time without access. The algorithm is
easy to realize and it makes full use of the known
information of the document. By taking contrast test with
LFU, SIZE
and LRU to access wind field data, result shows
FST algorithm’s advantages in terms of improving the
request hit rate, byte hit rate and reducing the delay time.
The first part of this paper analyzes the characteristics of
cache serving for wind power system. In the second part,
this paper studies traditional cache replacement algorithms
and proposes an algorithm FST based on comprehensive
factors. Next in the third part, the performance of the
replacement algorithms is tested according to the data
information of the wind power system. By comparing FST
algorithm with traditional algorithms in the test, result
proves the superiority of FST. The last chapter summarizes
the paper and illustrates the study result.
II. ANALYSIS OF CACHE FEATURE SERVICED IN WIND
POWER SYSTEM
A. Analysis of Access to Cache Object
Cache replacement algorithm is usually designed on the
basis of user’s access features, and there are mainly the
following three kinds of features.
171
2016 International Conference on Smart Grid and Clean Energy Technologies
978-1-4673-8904-
4/16/$31.00 ©2016 IEEE