Intel enters cloud computing to launch Hadoop CPU optimized version

Following Cloudera and Hortonworks, Intel at the end of February globally synchronized the launch of Intel-exclusive Hadoop distributed computing software, including the Hadoop release (Intel distribution) and the Hadoop management tool Intel Manager and Intel Active Turner, the first relevant software product since Intel invested heavily in research on cloud computing in 2009, Intel will enter the cloud data market with the Open-source distributed data analysis platform Hadoop.

Jason Fedder, general manager of Intel DCSG Asia Pacific, says the Hadoop release is based on Intel's hardware, providing a more secure, easier to deploy, and more efficient data processing platform for Hadoop users, and Intel will continue to invest in Hadoop development in the future. But Intel does not become a software company. Intel's goal is to create an open cloud data market ecosystem. So it will expand the Hadoop market through cross-industry alliances.

With the exception of OEM vendors, system integrators and independent software developers will be partners directly for the enterprise to deploy Intel version of Hadoop, and Intel provides educational training, Jason Fedder said. At present, with more than 20 hardware and software manufacturers are interested in alliances, including SAP, Cisco, SAS, 1degreenorth, Revolution Analytics and so on.

Intel's Hadoop proprietary software platform includes three main parts, the first is the management tools of the Hadoop platform, and the second is the special hardening based on the Hadoop framework, including HDFs, MapReduce, and HBase. At the same time and support SQL query instruction language hive, the query speed of SQL instruction increased 8.5 times times. Third, Intel's Hadoop architecture incorporates other common Hadoop tools such as zookeeper.

The Intel DCSG Asia Pacific Director, Patrick Buddenbaum, said that as Intel rewrote the Hadoop framework, it could directly support the Xeon advanced operation encryption instruction set AES-NI to improve data processing speed, and allow data collection, storage to the hbase process, can be protected by the chip's native encryption function, and performance is not affected by encryption. Patrick Buddenbaum says 1TB data processing can be shortened to 7 minutes with the Xeon E5-2690 processor, the Intel SSD 520 solid-state drive and the 10GbE network card.

Prior to this, Intel had launched Hibench as a reference indicator for deploying the Hadoop system's performance tuning. Patrick Buddenbaum said the current Intel's Hadoop software, which will provide commercial customers in subscription mode, was first deployed before the launch of the product, including China Mobile, Unicom, Nextbio and Intel. China Mobile, for example, deploys 133 Hadoop nodes to handle 30TB of monthly billing data.