论文部分内容阅读
The development of multi-core processors proves to be two trends: one is general purpose multi-core processor with low growth of core numbers,the other is many-core processors in data-intensive domain with rapid growth of core numbers.Nowadays and future on-line analytical processing shows the characteristics of big volume and high velocity,complex data management and data processing capacity are essential,the database management needs both the powerful data throughput and real-time analytical processing capacity with equal importance.This paper proposes co-OLAP technique on hybrid CPU and GPU platform with GPU oriented bitmap join index and GPU index processing.Distribution and query processing optimizations are implemented between CPU memory and GPU memory for the important multidimensional storage and index techniques,and with these approaches,different processors’ hardware advantages are fulfilled for improving big data latency between GPU memory and CPU memory.The overall performance for complex multidimensional analytical processing is improved for the hybrid CPU and GPU platform.