Data access analysis using entropy rate
A computer-readable non-transitory recording medium having stored therein a data analyzing program that causes a computer to execute a process. The process includes acquiring a transition matrix of data access based on a data access record of the data access; calculating an entropy rate for each of transition counts by using the transition matrix; determining mutual relevance of the data access based on the entropy rate; and storing data related to the data access on a disk, allocation of the data on the disk being determined by the mutual relevance of the data access.
1. A computer-readable non-transitory recording medium having stored therein a data analyzing program that causes a computer to execute a process, the process comprising:
acquiring a transition matrix of data access based on a data access record of the data access;
calculating an entropy rate for each of transition counts by using the transition matrix;
determining mutual relevance of the data access based on the entropy rate; and
storing data related to the data access on a disk, allocation of the data on the disk being determined by the mutual relevance of the data access,
wherein an effect of data locality is analyzed in accordance with a cache hit rate or a cache miss rate acquired based on the entropy rate and predetermined access patterns, and
the cache hit rate or the cache miss rate is calculated based on the transition counts where the entropy rate exceeds a logarithm of a predetermined segment size.
2. The computer-readable non-transitory recording medium according to claim 1 , wherein when N represents integers starting from 1 and indicating order numbers of a Markov chain representing the data access and the entropy rate is acquired for each value of N, the mutual relevance is estimated based on a change state of the entropy rate relative to N.
3. The computer-readable non-transitory recording medium according to claim 1 , wherein arrangement of data on the disk based on the data access is controlled based on the mutual relevance.
4. A method for analyzing data access used in a data access analysis apparatus, the method comprising:
acquiring a transition matrix of data access based on a data access record of the data access;
calculating an entropy rate for each of transition counts by using the transition matrix;
determining mutual relevance of the data access based on the entropy rate; and
storing data related to the data access on a disk, allocation of the data on the disk being determined by the mutual relevance of the data access,
wherein an effect of data locality is analyzed in accordance with a cache hit rate or a cache miss rate acquired based on the entropy rate and predetermined access patterns, and
the cache hit rate or the cache miss rate is calculated based on the transition counts where the entropy rate exceeds a logarithm of a predetermined segment size.
5. A data access analysis apparatus comprising:
a transition matrix generation unit configured to generate a transition matrix of data access based on a data access record of the data access;
an entropy rate calculation unit configured to calculate an entropy rate for each of transition counts by using the transition matrix; and
a locality analysis unit configured to determine mutual relevance of the data accesses based on the entropy rate and analyze locality by the mutual relevance of the data access,
wherein the locality analysis unit analyzes an effect of data locality in accordance with a cache hit rate or a cache miss rate acquired based on the entropy rate and predetermined access patterns, and
the locality analysis unit calculates the cache hit rate or the cache miss rate based on the transition counts where the entropy rate exceeds a logarithm of a predetermined segment size.