Research On The Query Processing Of Large-scale Product Knowledge In CPU+GPU Heterogeneous Environment

Under the background of the emergence and rapid development of e-commerce platforms,the number of goods is growing rapidly.Tens of thousands of users have commented on goods on shopping websites,and the number of comments on some popular goods is even more incalculable.These comments have important reference value for users.How to efficiently manage the massive product knowledge on the Internet is a hot issue in the current academic and industrial circles.At the same time,with the continuous development of Graphics Processing Unit(GPU)and other hardware,the heterogeneous mode system based on CPU + GPU has gradually become a research hotspot in the field of massive data management,and the CPU + GPU heterogeneous computing system plays an indispensable role in the future high-performance data management system.Based on large-scale product knowledge,this paper systematically studies the query processing technology of large-scale product knowledge in CPU + GPU heterogeneous environment:(1)In the existing work,this paper constructes the knowledge map of goods around the product information on the e-commerce platform,and proposed a knowledge representation method based on sparse matrix.The existing knowledge map of product is to construct the objective knowledge of product,but it does not involve the subjective knowledge such as the user's point of view,and ignores the potential value of the subjective information such as the user's point of view related to product.Therefore,this paper proposes a framework of product knowledge mapping,which combines objective knowledge and subjective knowledge of product,to realize the effective organization and management of objective knowledge and subjective knowledge of product.In addition,in order to solve the problem that product data are mostly represented by text,which leads to poor data retrieval efficiency,this paper proposed a knowledge representation method based on sparse matrix.Experimental results show that the sparse matrix knowledge representation method can reasonably use heterogeneous system resources and improve the system storage space utilization and data query efficiency.(2)This paper proposes a pipeline query optimization strategy in CPU + GPU heterogeneous environment.There are two kinds of knowledge management methods based on relational model and graph model,but with the exponential growth of product data,the retrieval efficiency can't satisfy the needs of users.At the same time,the continuous development of hardware technology makes the GPU function gradually expand from the graphics field to the general computing field.Compared with CPU,GPU is better at dealing with repetitive,simple logic and large-scale computing tasks.Aiming at the problem of poor efficiency of product knowledge retrieval,this paper adoptes GPU to speed up query processing,and proposes a pipeline product knowledge query optimization strategy under CPU + GPU heterogeneous environment.By reusing the query intermediate results,it makes full use of the powerful computing power of GPU.The experimental results show that the query performance of this method is significantly improved when dealing with user product knowledge retrieval.