Existing SQL level solutions are usually not scalable, so when the data becomes large when there may be obstacles. But this case, as has been the emergence of NoSQL great degree of ease, and with NoSQL technology improvement and maturity, this situation will be solved fundamentally.
We know that different from the relational database NoSQL thing is NoSQL does not use SQL as a query language, SQL interface provided as to why the NoSQL data storage HBase, the following reasons:
1. Use language such as SQL so easy to understand, so that people can more easily use HBase.
2. Use SQL such as a higher level language to write, reducing the amount of code written.
3. When the query is executed, data access and SQL runtime plus a layer of abstraction can be so heavily optimized between execution. For example, for the GROUP BY query, use HBase in co-processor, the polymerization can be carried out on the server, rather than on the client side, doing so would greatly reduce the amount of data transferred between the client and the server. Alternatively, you can execute in parallel GROUP BY client, which is based on the scope of Xingjian to truncate scanning achieved. The results will be returned via the parallel execution faster. All of these optimization without user involvement, you can simply execute the query.
Based on HBase SQL engine to achieve
There are some in the industry at this stage to try on HBase SQL engine layer, there are already some of the more stable and realistic solutions.
1.Hive integration HBase
Hive and HBase integration feature from Hive0.6.0 version has begun to appear, the use of both external API interface to communicate with each other, communication is mainly rely hive_hbase-handler.jar Kit (Hive Storage Handlers). Because there is a relatively large HBase version changes, so not every version of the existing HBase and Hive can integrate version, so in the course special attention is consistency between the two versions.
Phoenix Salesforce.com by the open source Apache HBase is built on top of a SQL intermediate layer that allows developers to execute SQL queries on HBase. Phoenix is completely written in Java code is on Github, and provides a client can be embedded JDBC driver. For simple queries 10w to 100w line is, Phoenix be better than Hive.
Kundera is a JPA2.0 object mapping framework compatible with NoSQL data stores. Kundera build on existing library, package the simple API, its main characteristics are:
1) support cross persistent data storage, which means that different users can use a single method of data storage to store and retrieve related entities.
2) can be well-managed affairs, while supporting EntityTransaction and Java Transaction API (JPA).
3) Compatible JPA2.0, strictly use JPA annotation object to the map data storage table.
4) currently supports NoSQL servers include: HBase, MongoDB, Redis, Neo4j like.
There are other solutions, such as: Lealone, hbase-sql, Impala, etc., or immature, or stop updating it, or has limitations. The readers are interested in, you can get to know yourself.