Moving Towards Pure ANSI SQL in NoSQL

薛瑞夫(Sheriffo Ceesay)

指導教授 ： 鍾葉清

國立清華大學/電機資訊學院/資訊系統與應用研究所/碩士(2013年)

摘要

Moving Towards Pure ANSI SQL in NoSQL
The main focus of this master’s thesis is to narrow down the user friendly gap between
the newly more distributed data processing platforms (HBase,Cassandra,
MapReduce e.t.c) and the traditional less distributed data processing platforms
e.g. (RDBMS’s).
Lot of work have been done in this area e.g. Hive and Pig but they are not
pure SQL.
Over the past few decades RDBMS’s and data-warehouses were the only choice
of data processing platforms with rich set of data processing tools e.g. SQL but
recently, due to the variety, velocity and volume of data, these traditional data
processing platforms becomes less efficient to handle this kind of data; thereby
the need to come up with more efficient data stores and processing platforms.
Though NoSQL data stores have lived up to their expectations of storing and
processing large datasets but this process might not be simple and convenient
as in traditional databases. One common cons of NoSQL databases is the lack
of the much loved SQL language.
This thesis will therefore focus on this new type of data stores also called
(NoSQL). Specifically we will focus on HBase which is a column oriented or
BigTable like Database as our choice of NoSQL store.
The fact that NoSQL databases are becoming very popular we will propose our
data mapping methods which can help migration from Relational Databases to
NoSQL databases to be less daunting.
Since this movement is from RDB’s which has rich set of procedures i.e. SQL to
access and manipulate data, we will extend our work to bridge the gap between
SQL and NoSQL by providing methods of using pure ANSI SQL to manipulate
the underlying data which is stored in our NoSQL store (HBase).

關鍵字

雲端運算 ； 軟體框架 ； 資料庫 ； 數據庫 ； 資料庫 ； 數據庫查詢語言

並列摘要

Moving Towards Pure ANSI SQL in NoSQL
The main focus of this master’s thesis is to narrow down the user friendly gap between
the newly more distributed data processing platforms (HBase,Cassandra,
MapReduce e.t.c) and the traditional less distributed data processing platforms
e.g. (RDBMS’s).
Lot of work have been done in this area e.g. Hive and Pig but they are not
pure SQL.
Over the past few decades RDBMS’s and data-warehouses were the only choice
of data processing platforms with rich set of data processing tools e.g. SQL but
recently, due to the variety, velocity and volume of data, these traditional data
processing platforms becomes less efficient to handle this kind of data; thereby
the need to come up with more efficient data stores and processing platforms.
Though NoSQL data stores have lived up to their expectations of storing and
processing large datasets but this process might not be simple and convenient
as in traditional databases. One common cons of NoSQL databases is the lack
of the much loved SQL language.
This thesis will therefore focus on this new type of data stores also called
(NoSQL). Specifically we will focus on HBase which is a column oriented or
BigTable like Database as our choice of NoSQL store.
The fact that NoSQL databases are becoming very popular we will propose our
data mapping methods which can help migration from Relational Databases to
NoSQL databases to be less daunting.
Since this movement is from RDB’s which has rich set of procedures i.e. SQL to
access and manipulate data, we will extend our work to bridge the gap between
SQL and NoSQL by providing methods of using pure ANSI SQL to manipulate
the underlying data which is stored in our NoSQL store (HBase).

並列關鍵字

Hadoop ； MapReduce ； NoSQL ； Hive ； HBase ； ANSI SQL

參考文獻

[1] David Dewitt and Jim Gray, Parallel Database System: The future of High

Performance Database Systems, ACM 1992.

for Structured Data

[5] Jeffrey Dean and Sanjay Ghemawat. MapReduce: Simplified Data Processing

[8] Chongxin Li. RDB to HBase: Transforming Relational Database into