Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
en:services:application_services:high_performance_computing:spark [2019/08/09 13:30]
ckoehle2 created
en:services:application_services:high_performance_computing:spark [2019/09/05 16:49] (current)
ckoehle2 link to user wiki
Line 1: Line 1:
 ====== Apache Spark ====== ====== Apache Spark ======
  
-//under construction//+===== Introduction ===== 
 +[[https://spark.apache.org/​|Apache Spark]] is a distributed general-purpose cluster computing system. 
 + 
 +Instead of the classic Map Reduce Pipeline, Spark’s central concept is a resilient distributed dataset (RDD) which is operated on with the help of a central driver program making use of the parallel operations and the scheduling and I/O facilities which Spark provides. Transformations on the RDD are executed by the worker nodes in the Spark cluster. The dataset is resilient because Spark automatically handles failures in the Worker nodes by [[https://​spark.apache.org/​docs/​latest/​rdd-programming-guide.html|redistributing]] the work to other nodes. 
 + 
 +In the following sections, we give a short introduction on how to prepare a Spark cluster and run applications on it in the Scientific Compute Cluster. 
 +===== Creating a Spark Cluster on the SCC ===== 
 +<WRAP center round important 60%> 
 +We assume that you have access to the HPC system already and are logged in to one of the frontend nodes ''​gwdu101'',​ ''​gwdu102''​ and ''​gwdu103''​.\\ If that's not the case, please check out our [[en:​services:​application_services:​high_performance_computing:​running_jobs_slurm|introductory documentation]] first. 
 +</​WRAP>​ 
 + 
 +Apache Spark is installed in version 2.4.3, the most recent stable release at the time of this writing. The shell environment is prepared by loading the module ''​spark/​2.4.3'':​ 
 + 
 +<​code>​ 
 +gwdu102 ~ > module load spark/​2.4.3 
 +</​code>​ 
 + 
 +We’re now ready to deploy a Spark cluster. Since the resources of the HPC system are managed by [[en:​services:​application_services:​high_performance_computing:​running_jobs_slurm|Slurm]],​ the entire setup has to be submitted as a job. This can be conveniently done by running the script ''​scc_spark_deploy.sh'',​ which accepts the same arguments as the sbatch command used to submit generic batch jobs: 
 + 
 +<​code>​ 
 +gwdu102 ~ > scc_spark_deploy.sh —nodes=2 —time=02:​00:​00 
 +Submitted batch job 872699 
 +</​code>​ 
 + 
 +In this case, the ''​%%--%%nodes''​ parameter has been set to specify a total amount of two worker nodes and ''​%%--%%time''​ is used to request a job runtime of two hours. The job ID is reported back. We can use it to inspect if the job is running yet and if so, on which nodes: 
 + 
 +<​code>​ 
 +gwdu102 ~ > squeue —jobs=872699 
 +             JOBID PARTITION ​    ​NAME ​    USER ST       ​TIME ​ NODES NODELIST(REASON) 
 +            872699 ​  ​fat-fas ​   Spark ckoehle2 ​ R       ​1:​59 ​     2 dfa[008-009] 
 +</​code>​ 
 + 
 +The first node reported in the //​NODELIST//​ column is running the Spark master. Its hostname is used to form a URL like ''​spark:​%%//​%%host:​port''​ that the spark applications,​ such as ''​spark-submit''​ and ''​spark-shell''​ need to connect to the master: 
 + 
 +<​code>​ 
 +gwdu102 ~ > spark-shell —master spark://​dfa008:​7077 
 +</​code>​ 
 + 
 +Here, the Spark shell is started on the frontend node ''​gwdu102''​ and connects to the master ''​dfa008''​ on the default port ''​7077''​. 
 +{{ :​en:​services:​application_services:​high_performance_computing:​spark:​shell_setup.png?​nolink&​800 |}} 
 +Scala code that is entered in this shell and parallelized with Spark will be automatically distributed across all nodes that have been requested initially. N.B.: The port that the application’s web interface is listening on (port ''​4040''​ by default) is also being reported in the startup message. 
 + 
 +Once the Spark cluster is not needed anymore, it can be shut down gracefully by using the provided script ''​scc_spark_shutdown.sh''​ and specifying the job ID as an argument: 
 + 
 +<​code>​ 
 +gwdu102 ~ > scc_spark_shutdown.sh 872699 
 +</​code>​ 
 + 
 +In case a single node is sufficient, Spark applications can be started inside a Slurm job without previous cluster setup - the ''​%%--%%master''​ parameter can be omitted in that case. If you want to quickly test your application on a frontend node or inside an interactive job, this approach is not feasible since by default all available CPU cores are utilized, which would disturb the work other users of the system. However, you can specify the URL ''​local[CORES]'',​ where //CORES// is the amount of cores that the Spark application should utilize to limit your impact on the local system, for example: 
 + 
 +<​code>​ 
 +gwdu102 ~ > spark-shell —master local[4] 
 +</​code>​ 
 + 
 +===== Access and Monitoring ===== 
 +Once your Spark cluster is running, information about the master and workers is being printed to the file ''​scc_spark_job-$JOBID.out''​ in the current working directory you deployed the cluster from. For example, in the case at hand, the MasterUI, a built in web interface that allows us to check the master for connected workers, the resources they provide as well as running applications and the resources they consume, is listening on the master on port ''​8082''​. 
 + 
 +{{ :​en:​services:​application_services:​high_performance_computing:​spark:​master_webui.png?​nolink&​800 |}} 
 + 
 +Inside [[en:​services:​network_services:​goenet:​start|GöNET]],​ an SSH tunnel allows us to open the web interface on ''​localhost:​8080''​ by starting OpenSSH with the following parameters:​ 
 + 
 +<​code>​ 
 +ssh -N -L 8080:​dfa008:​8082 -l ckoehle2 gwdu102.gwdg.de 
 +</​code>​ 
 + 
 +===== Example: Approximating Pi ===== 
 + 
 +To showcase the capabilities of the Spark cluster set up thus far we enter a short [[https://​spark.apache.org/​examples.html|Scala program]] into the shell we’ve started before. 
 + 
 +{{ :​en:​services:​application_services:​high_performance_computing:​spark:​shell_example.png?​nolink&​800 |}} 
 + 
 +The local dataset containing the integers from //1// to //1E9// is distributed across the executors using the parallelize function and filtered according to the rule that the random point //(x,y)// with //0 < x, y < 1// that is being sampled according to a uniform distribution,​ is inside the unit circle. Consequently,​ the ratio of the points conforming to this rule to the total number of points approximates the area of one quarter of the unit circle and allows us to extract an estimate for the number //Pi// in the last line. 
 + 
 +===== Further reading ===== 
 +You can find a more in-depth tour on the Spark architecture,​ features and examples (based on Scala) in the [[https://​info.gwdg.de/​wiki/​doku.php?​id=wiki:​hpc:​parallel_processing_with_spark_on_gwdg_s_scientific_compute_cluster|HPC wiki]]. 
 + 
 + --- //​[[christian.koehler@gwdg.de|ckoehle2]] 2019/09/02 19:50//