Cloudera moves Spark into Hadoop

http://www.aliyun.com/zixun/aggregation/13383.html"> Spark memory computing framework for a variety of iterative algorithms and interactive data analysis to improve the real-time processing of big data and accuracy.And MapReduce processing framework Is good at complex batch operations, login filtering, ETL (data extraction, conversion, loading), web indexing and other applications, MapReduce has been criticized for low latency business.

Figure 1: Spark memory computing framework

"Spark records every operation that data produces and can reliably store that data in memory, making it ideal for cover-up calculations and efficient iteration algorithms," said Cloudera.

It is reported that Cloudera Enterprise Data Hub version provides unlimited support for a variety of advanced components, such as Interactive SQL Analysis Impala, interactive search, navigation data management and Hbase NoSQL. Enterprise Flex versions provide optional component versions, while Enterprise Basic versions provide only the core Hadoop components.

According to Cloudera, Cloudera will install the Spark component in both releases. Users can take advantage of it as an optional component of the Enterprise Flex release or as a component of the Enterprise Data Hub release.

Figure 2: Spark On YARN

It is reported that the product uses Spark 0.9.0, Spark standalone mode has been tested in Cloudera Enterprise Data Hub4.4.0. In the near future, Cloudera said Spark will also be supported in Enterprise 5.0 and YARN.

Editors: 1.Hadoop vs Spark performance comparison

2. Youku potatoes use Spark to improve big data applications