FAST FLASH MEMORY CACHING BASED ON FILE ACCESS FREQUENCY
Author(s):
Chenhan Liao,
Frank Wang,
Na Helian,
Sining Wu,
Yuhui Deng
Paper abstract: In modern file systems, traces monitor the file operations and user behaviors. Inevitably, large amount
of data continuously produced in daily manner. We show that the knowledge hidden behind system
traces can help us understand the system and user behaviors. In this paper, we illustrate that once a file is
created with a set of attributes, such as name, type, permission mode, owner and owner group, its future
access frequency is predictable. A decision-tree-based predictive model is established to predict whether
a file will be frequently accessed or not. By consulting with the rules generated from the predictive
model over diverse real-system NFS traces, the model can predict a newly created file’s future access
frequency with sufficient accuracy. We further introduce an evolutionary storage system, which employs
the predicted frequency information to decide what files to keep in a fast storage device, flash memory.
The trace-driven experimental results indicate that the performance speedup due to the predictionenabled
optimization is 2-4 compared with base case.
