HBase Seven years--bdtc2013 lecturer Michael Stack

Engineer,michael Stack LinkedIn has the largest number of job descriptions in his CV, and "Engineer" is the data-field veteran's evaluation of himself. When we learn more about Michael, we find that he does not need words like "Leader" or "Senior".

As a native database of Hadoop, HBase is widely found in the architecture of large data analysis system. However, the official Apache team of engineers is still less than 40, Michael is one of them, with more than 1 ways to submit the code directly. At the same time, Michael Stack is now the chairman of the HBase Project Management committee, which has been directing the development of HBase since 2007.

Michael's LinkedIn engineer career began in 1988, during which he worked for a number of well-known companies such as Microsoft, and in October 2012 Michael joined Cloudera.

Cloudera & Impala

At the end of October 2012, Cloudrea open-source A real-time query project Impala based on Hadoop. Cloudera by Facebook, Google and Yahoo! 's former engineers Jeff Hammerbacher, Christophe Bisciglia, Amr Awadallah, and the incumbent CEO, Oracle former executive Mike Olson was created, Impala was released when the company was just 4 years old.

Impala is based on the Apache Drill project development, while Apache drill is Google Dremel's evolutionary product. As we all know, Google has mapreduce and its derivatives caffeine on large data processing, and the search giant can spend a lot of effort to develop Dremel. So it's no surprise that Impala is 3~90 times faster than the hive SQL query based on MapReduce.

HBase's internal and internal

On the descent, HBase was born noble Google BigTable's Open source realization, and in the promotion, HBase is also the dominant--natural integration with Hadoop. In the NoSQL realm, however, the list of store-popularity kings was cassandra--by Facebook. Although the social giants abandoned Cassandra to HBase in a few years, many of the problems in HBase still constrain its development, such as strong Java-featured APIs, complex operations, failover, batch file system HDFs, and so on. "Worse" is that Hadoop reconstructed the MapReduce framework in version 2.0, and the new version yarn borrowed from the Mesos feature, proposing container the resource isolation framework, allowing more frameworks to run on the Hadoop cluster.

"Internal and internal internal and internal" let HBase Summit NoSQL Road More Misty, fortunately, the first China large data technology conference, we had the honor to invite the HBase Project Management Committee chairman Michael Stack for us to analyze the future of hbase and the status quo, To share his first-hand practice on the HBase project, please pay attention to the CSDN follow-up report for more details.