Date Added: Dec 2008
As the price of memory continuously becomes cheaper, many of database tables and indices are easily placed in main memory. Traditional bottleneck of disk access is replaced by memory access. It has been revealed that commercial DBMSs spend half the execution time on memory access when a whole database fits in memory. As a result, today's database applications are becoming increasingly compute and memory bound. Many works address the techniques focusing on minimizing memory latencies when the database is accessed. The latency time in retrieving data from the memory resided with a huge index is considerably time-consuming. In fact, compactly coordinating the data arrangement in a database index is essential toward the acquisition of memory performance.