The Method Of Flash Based Mass Non-relation Storage

With the development of the internet industry, data increasingly appear indifferent kinds of industries, which greatly promote the progress of the society.Furthermore, various techniques arise at the historic moment on the basis of theincrease and application of mass data. On the other hand, the application of thehardware such as SSD makes the data achieve great improvement of theperformance in the application field.The application of mass data is different from relational data models in thetraditional database, which typically is the application of non-relevant database indistributed areas, such as the framework development of hadoop. However, thecommon system structure is set on the basis of traditional hardware withoutconsideration of the characteristics of hardware such as SSD, therefore,performance optimization basically focuses on the node communication, loadbalance etc., which ignores the development of hardware. Consequently, how toorganically bond mass data, non-relevant data structure, SSD in together,optimizing the existing models, and improving reading and writing capability arethe key points in this paper.This paper proposes to improve the performance of reading and writing basedon the common system structure of flash memory, which specifically shows asfollows:In writing, changes the process of random writing into the process ofcontinuous writing through the reasonable data structure and algorithm research.Moreover, improves the writing algorithm by comparison of the performance ofSSD continuous writing and random writing, as well as the performance of SSD andordinary hard disks. Lastly, does the research of writing optimization algorithm onthe basis of full consideration and use of hardware characteristics of SSD.In reading, proposes the multiple bloom filter research based on the SSDthrough the improvement of the structure of bloom filter. Furthermore, changes thetraditional bloom filter from one inquiring process to the step inquiring processthrough the transformation for the traditional bloom filter, which helps to achievethe improvement of the performance. Besides, makes better the data structure andtakes the advantage of characteristics such as reading fast of SSD and high cost of bloom filter false positive inquires to improve the performance of the reading.