At Linaro Connect Budapest 2017, ARM and the Linaro LEG (Linaro Enterprise Group) team demonstrate Apache Ambari running on ARM servers. Apache Ambari is one of key Big Data components that provides an easy to use web interface to provision, manage and monitor Hadoop cluster and various other Big Data tools. In this demo Apache Ambari is running on 3 node cluster with Hadoop, YARN and Zookeeper, all on AARCH64 hardware.
With ARM servers getting into Datacenter, Linaro has been collaborating with ARM and ARM vendors in making sure Big data components work well in AARCH64 architecture. Porting and Building Apache Ambari on AARCH64 is one of the efforts the team has been working on apart from porting, building and benchmarking Hadoop, Spark, Hive, HBase and other Big data components. The team chose to showcase Apache Ambari as a high level component due to it being very intuitive, easy-to-use Hadoop management web UI backed by its RESTful APIs. Ambari provides a dashboard with metrics for CPU, Storage, memory utilization and also metrics for HDFS, MapReduce, Pig, Hive etc for monitoring Hadoop Cluster. It also provides step-by-step wizard for managing Hadoop Clusters (like adding nodes, taking down nodes, doing rolling upgrades, etc).
As of today, Ambari is officially supported only on X86 servers. The work team has done is to patch Ambari to work on AARCH64, which involved patching various dependencies like phantomjs, leveldb, leveldbjni java libaries, etc. Linaro is also part of ODPi organization (odpi.org), which has been working on standardizing Big Data. Ambari is part of ODPi’s operations specs. The collaboration helps in speeding up upstreaming process since ODPi also has some of same maintainers as of ASF.
The work done for this demo is a PoC running in ARM lab (working on Linaro Dev Cloud at the mean time) and yet published.
This video features Eugene Xie ARM Principal Software Engineer & Tech lead of Workloads team of Enterprise Software, BSG, Ganesh Raju – Tech Lead, BigData team and Naresh Bhat, Cavium assignee to Linaro for BigData.