Research On Retrieval Speed Improvement Of HBase Based On Coprocessor Mechanism

In the fie ld of big data processing,the open source framework of Hadoop and HBase is more and more wide ly used.HBase,as a distr ib uted column-oriented high-volume database,is used in t he storage and computat io n of many struct ured data to replace the traditional database as they couldn't cope with the greater amount of data.Due to the nature of HBase distr ibuted,column or iented storage,mak ing it diffic ult for HBase itse lf to build the non-primary key index(secondary indexes)like a tradit io na l database.HBase data records are ident ified by Rowkey uniq ue,and we can access to the data only through Rowkey.In order to improve the retrie va l speed of HBase,the research work of HBase secondary indexes are wide ly carr ied out.Restricted to the server-side funct io nal exte nsio n provided by the HBase are very few,the progress of early H Base indexing research work is ver y sma ll.Ear ly indexing schema can only build the index co mponent outs ide HBase cluster,it makes the index separated from data,and the index is ine ffic ient.As the HBase coprocessor mecha nis m is put forward,the status quo to improve.Coprocessor mechanis m allows clients to call ser ver-side code remotely and set the server processing logic.So that it makes the indexing work migrated to the server.So the data and the index can be built together and the indexing performance will be greatly improved.For the fact that HBase has no secondary index funct ion,and there are ma ny defects of var io us index sche mes,this paper stud ies deep ly HBase interna l mecha nis ms,and seek advantages and defects of var ious index scheme,the n proposed two kinds of indexing scheme based on coprocessor.1.This paper proposed an index tab le schema of HBase based on coprocessor.It use the coprocessor mecha nis m and t he HBase table to build an index tab le in the HBase cluster correspiond ing to the HBase da ta table and design t he logic of the update and query of index.It improved the process of tw ice table lookup,as to realize the index of HBase.2.Thie paper a lso proposed another index sc hema o f HBase based on coprocessor.To improve the sche ma of inde x table,it builds t he index in t he me mory object of coprocessor.Wit h the high read-write speed of the me mory,it improved the efficiency of the index of HBase and data retrieval speed further.This paper comp letes the design and imp le mentat ion o f the indexing sche me of HBase based on coprocessor mecha nis m.And ver ify the improve ment of H Base retrieva l speed wit h the indexing schema.It is proved that it is effect ive to build the HBase index with coprocessor mechanism to improve retrieval speed.