MMDB Reload Concerns

Abstract Due to the volatility of main memory, a Memory Database (MMDB) system may lose the primary copy of its database after a system failure. The lost database can be recovered by reloading the archive database residing on Archive Memory (AM) into Main Memory (MM) and applying log data. This paper analyzes three major issues that directly impact the reload performance: how to organize the database on AM, how to choose the reload granularity, and how to create address translation tables. The paper proposes an AM structure which makes use of disk striping to reduce reload time, and a special AM structure on which data blocks are organized based on their access frequency order to ensure hot spots are reloaded before less frequently used data. Our analysis shows that cylinder is the best reload granularity in terms of total reload time and transaction waiting time.

[1]  Douglas Comer,et al.  Operating system design: the XINU approach , 1984 .

[2]  Michael J. Carey,et al.  A recovery algorithm for a high-performance memory-resident database system , 1987, SIGMOD '87.

[3]  Le Gruenwald,et al.  Reload in a main memory database system: MARS , 1990, Proceedings of the 1990 Symposium on Applied Computing.

[4]  Vijay Kumar,et al.  Performance measurement of some main memory database recovery algorithms , 1991, [1991] Proceedings. Seventh International Conference on Data Engineering.

[5]  Margaret H. Eich.  Mars: The Design of a Main Memory Database Machine , 1988 .

[6]  Michelle Y. Kim.  Parallel Operation of Magnetic Disk Storage Devices: Synchronized Disk Interleaving , 1985, IWDM.

[7]  Ravi Krishnamurthy,et al.  Design of a Memory Resident DBMS , 1985, IEEE Computer Society International Conference.

[8]  Michael Stonebraker,et al.  Implementation techniques for main memory database systems , 1984, SIGMOD '84.

[9]  George Paul,et al.  Disk Interleaving and Very Large Fast Fourier Transforms , 1987 .

[10]  Hector Garcia-Molina,et al.  Disk striping , 1986, 1986 IEEE Second International Conference on Data Engineering.

[11]  Robert B. Hagmann.  A Crash Recovery Scheme for a Memory-Resident Database System , 1986, IEEE Transactions on Computers.

[12]  Gio Wiederhold.  File organization for database design , 1987 .

[13]  Michelle Y. Kim,et al.  Synchronized Disk Interleaving , 1986, IEEE Transactions on Computers.

[14]  Hector Garcia-Molina,et al.  System M: A Transaction Processing Testbed for Memory Resident Data , 1990, IEEE Trans. Knowl. Data Eng..

[15]  Le Gruenwald,et al.  MMDB reload algorithms , 1991, SIGMOD '91.