HSQL: 具高擴展性的雲端線上交易處理資料庫系統

Abstract

傳統的關聯式資料庫系統遇到巨量資料時，會有擴展性的問題，為了能夠有效率 的處理這些巨量的資料，許多NoSQL的資料庫已經被開發出來，然而，許多的功能，像是SQL的介面、多個列的交易處理以及次級索引，在NoSQL的資料庫裡並不支援。因此，在NoSQL的資料庫裡，提供這些功能已經變成一個重要的研究議題。 在這篇論文中，我們發展了HSQL，一個具有高擴展性的OLTP資料庫系統。HSQL使用HBase當作底層的分散式儲存系統，因此具有HBase的高擴展性。為了能夠處理OLTP的工作，我們提出了一個創新的方法，在HBase上實作多個列的交易處理。另外，我們也在HBase上設計了一個分散式的次級索引，我們實驗結果顯示我們的資料庫系統在處理巨量資料時，相較於MySQL有較好的性能。

With the ever-increasing large amounts of data, traditional relational database management systems may suffer limited scalability. To be able to process large amounts of data, many NoSQL databases have been developed. However, many features, such as SQL interface, multi-row transactions,and secondary index support, are unavailable in the NoSQL databases. Providing these features for NoSQL databases has become an important research issue. In these thesis, we develop HSQL, a highly scalable database for OLTP applications. HSQL uses HBase as the underlying distributed data store and interits the scalability of HBase. To be able to process OLTP workload, we propose a novel approach to support multi-row transactions on HBase. In addition, we also devise a distributed secondary index scheme for HBase. The experiment results show that HSQL has better performances than MySQL on large scale of data.