HARBOR - High Availability and Replication-Based Online Recovery
My fifth year at MIT culminated in my Master of Engineering thesis, completed under research advisor Professor Sam Madden. For my thesis, I designed, implemented, and evaluated HARBOR -- a new logless crash recovery and high availability framework targeted toward distributed data warehouses. My research nabbed me the 2006 Chorafas Foundation Award.
Any highly available database system will use data redundancy so that the system can continue to service queries despite site failures. The key idea of HARBOR is to capitalize on the data redundancy to perform crash recovery via remote querying of live sites rather than via processing of a local log. This logless approach to recovery is simpler and faster than traditional log-based approaches, and it avoids the runtime overhead of synchronous log writes during commit processing.
For further information, contact me or consult the following resources:
- Edmond Lau, Samuel Madden, "An Integrated Approach to Recovery and High Availability in an Updatable, Distributed Data Warehouse", VLDB 2006 [pdf]
- Edmond Lau, "HARBOR: An Integrated Approach to Recovery and High Availability in an Updatable, Distributed Data Warehouse," Master's Thesis [pdf]
- Edmond Lau, "A Safe HARBOR for Data Warehouses," MIT EECS MasterWorks presentation (delivered 4/27/06) [ppt]
- Edmond Lau, "A Safe HARBOR for Data Warehouses," Database Group presentation (delivered 5/12/06) [ppt]