Making B ⁺ -tree efficient in PCM-based main memory

Abstract

Phase change memory (PCM) is a promising technology for building future large-scale and low-power main memory systems. Main memory databases (MMDBs) can benefit from the high density of PCM. However, its long write latency, high write energy, and limited lifetime, bring challenges to database algorithm design for PCM-based memory systems. In this paper, we focus on making B + -tree PCM-friendly by reducing the write accesses to PCM. We propose three different schemes. Experimental results show that they can efficiently improve the performance, reduce the memory energy consumption, and improve the lifetime for PCM memory.

On the analysis of big data indexing execution strategies

“…Following is the definition of Index Hit Ratio with the condition that selection predicates of incoming queries are normally distributed: (4) where ,…”

“…Index rebuilding is easy [4]. It requires to perform delete operation on previous index and iterate create index operation.…”

SmallClient for big data: an indexing framework towards fast data retrieval

“…Furthermore, the tree based structure makes it fast in traversing as compared to other non-clustered indexes. Recent implementations of B-Tree indexing are on flash memory optimization [26] and on main memory [27].Apache Lucene [34] implements inverted indexes that are non-clustered indexes on big data to retrieve required data by searching indexes instead of whole data sets. However, the index design implemented by Apache Lucene has few limitations: indexing on whole data set requires a significant size of main memory, searching data for nonindexed attributes and searching whole row when selective attributes are indexed is not available with Apache Lucene indexes.…”

Efficient Tree Indexing for PCM-Based Memory Systems

“…The properties of PCM lie on the gap between DRAM and NAND flash [6]. PCM has following advantages: (1) it has high storage density, which is about 2 to 10 times higher than DRAM, that means it can store more information in less space, and improve the efficiency of many database applications; (2) the energy consumption of PCM is significantly decreased, and the idle energy consumption is almost 0; (3) read latency is almost the same with DRAM; (4) it's non-volatile characteristic makes data persistent.…”

“…Unsorted scheme [6,12] is helpful in reduce write count to keep nodes in-order, but it cannot avoid sort operation when node split and merge. Overflow node scheme [6,12] is good in write operations because the split and merge operations are given priority within the overflow chain, thereby reduce the write operation to the parent node. However, once node size is small (less than 4 cacheline_size) [6], it does not have performance improve.…”