A design for high-performance flash disks
ACM SIGOPS Operating Systems Review - Systems work at Microsoft Research
Design of flash-based DBMS: an in-page logging approach
Proceedings of the 2007 ACM SIGMOD international conference on Management of data
Design tradeoffs for SSD performance
ATC'08 USENIX 2008 Annual Technical Conference on Annual Technical Conference
Flash Disk Opportunity for Server Applications
Queue - Enterprise Flash Storage
Online maintenance of very large random samples on flash storage
Proceedings of the VLDB Endowment
FlashLogging: exploiting flash devices for synchronous logging performance
Proceedings of the 2009 ACM SIGMOD International Conference on Management of data
Advances in flash memory SSD technology for enterprise database applications
Proceedings of the 2009 ACM SIGMOD International Conference on Management of data
Evaluating and repairing write performance on flash devices
Proceedings of the Fifth International Workshop on Data Management on New Hardware
Evaluating Non-In-Place Update Techniques for Flash-Based Transaction Processing Systems
DEXA '09 Proceedings of the 20th International Conference on Database and Expert Systems Applications
Disk schedulers for solid state drivers
EMSOFT '09 Proceedings of the seventh ACM international conference on Embedded software
RS-Wrapper: random write optimization for solid state drive
Proceedings of the 18th ACM conference on Information and knowledge management
Proceedings of the 2010 ACM SIGMOD International Conference on Management of data
Hi-index | 0.00 |
Flash memories are considered a competitive alternative to rotating disks as non-volatile data storage for database management systems. However, even if the Flash Translation Layer - or FTL - allows both technologies to share the same block interface, they have different preferred access patterns. Database management systems could potentially benefit from flash memories as they provide fast random access for read operations although random writes are generally not as efficient as sequential writes. In this paper, we propose a simple data placement algorithm designed for flash memories, to reorganize inefficient random writes in a quasi-sequential access pattern. This access pattern is first established encouraging for a subset of flash devices by identifying a strong correlation between spatial locality and write performances, with a distance being defined to quantify this effect. This design is then validated by a formalization with a mathematical model, along with experimental results. With this optimization, random write potentially become as efficient as sequential write, improving random write speed by up to two orders of magnitude.