BigData Facility

The BigData Facility (DHYAN) has 20 CPU (800 cores) data node clusters with 8TB of RAM. The BigData Facility use Parallel File System (~250TB) of DDN GRIDScaler storage at 15 GBps throughput over 100 Gbps interconnect network.

Submitting job on BigData Cluster:

• Apache Spark

Spark built on the concept of Distributed Datasets, which contain arbitrary Java or Python objects. You create a dataset from external data, and then apply parallel operations to it. The building block of the Spark API is its RDD API. In the RDD API, there are two types of operations: transformations, which define a new dataset based on previous ones, and actions, which kick off a job to execute on a cluster. On top of Spark’s RDD API, high-level APIs are provided, e.g. DataFrame API and Machine Learning API. These high-level APIs provide a concise way to conduct certain data operations. In this page, we will show examples using RDD API as well as examples using high level APIs.

Note: Your files or directory path defined in hdfs path

Check HDFS file system: hdfs dfs –ls username

Spark examples

      spark-submit
    --class org.apache.spark.examples.SparkPi
    --master yarn
    --deploy-mode cluster
    --driver-memory 4g
    --executor-memory 2g
    --executor-cores 1
    --queue default
    spark-examples_2.11-2.3.0.2.6.5.0-292.jar 10



Useful Commnads

      hdfs dfs -mkdir DIR_PATH
      hdfs dfs -rm r DIR_PATH / File
      hdfs dfs -put SOURCE_PATH DEST_PATH
      hdfs dfs -get SOURCE_PATH DEST_PATH


• Data Frame API

In Spark, a DataFr ame is a distributed collection of data organized into named columns. Users can use DataFrame API to perform various relational operations on both external data sources and Spark’s built in distributed collections without providing specific procedures for processing data. Also, programs based on DataFrame API will be automatically optimized by Spark’s built in optimizer, Catalyst. Examples:

Text Search: In this example, we search through the error messages in a log file. Written in python)

      textFile = sc.textFile("hdfs://...")
# Creates a DataFrame having a single column named "
      df = textFile.map(lambda r: Row(r)).toDF(["line"])
      errors = df.filter(col("line").like("%ERROR%"))
# Counts all the errors
      errors.count()
# Counts errors mentioning MySQL
      errors.filter(col("line").like("%MySQL%")).count()
# Fetches the MySQL errors as an array of strings
      errors.filter(col("line").like("%MySQL%")).collect()



Simple Data Operations: In this example, we read a table stored in a database and calculate the number of people for every age. Finally, we save the calculated result to S3 in the format of JSON. A simple MySQL table "people" is used in the example and this table has two columns, "name" and "age". Examples: (Written in python)

# Creates a DataFrame based on a table named "people"
# stored in a MySQL
      url = "jdbc:mysql://yourIP:yourPort/test?user=yourUsername;password=yourPassword"
      df = sqlContext \
      .read \
      .format("jdbc") \
      .option("url", url) \
      .option("dbtable", "people") \
      .load()
# Looks the schema of this DataFrame.# Looks the schema of this DataFrame.
      df.printSchema()df.printSchema()
# Counts people by age# Counts people by age
      countsByAge = df.groupBy("age").count()
      countsByAge.show()
# Saves countsByAge to S3 in the JSON format.# Saves countsByAge to S3 in the JSON format.
      countsByAge.write.format("json").save("s3a://...")


• Machine learning Example

MLlib, Spark’s Machine Learning (ML) library, provides many distributed ML algorithms. These algorithms cover tasks such as feature extracti on, classification, regression, clustering, recommendation, and more. MLlib also provides tools such as ML Pipelines for building workflows, CrossValidator for tuning parameters, and model persistence for saving and loading models.

Prediction with Logistic Regression: In this example, we take a dataset of labels and feature vectors. We learn to predict the labels from feature vectors using the Logistic Regression algorithm. Examples:

# Every record of this DataFrame contains the label and features represented by a vector.
      df = sqlContext.createDataFrame(data, ["label", "features"])
# We limit the number of iterations to
      lr = LogisticRegression(maxIter=10)
# Fit the model
      model = lr.fit(df)
# Given a dataset, predict each point's label, and show
      model.transform(df).show()

 

Useful Commands

• Accessing a user account: ssh <username>@172.20.70.16

 

Usage Guidelines

• Users are not supposed to run any job on the login node.
• Users are not allowed to run a job by directly login to any data node.
• Users must report for any weaknesses in computer security and incidents of possible misuse or violation of the account policies to the HPC administrators or write regarding the same to ccf@iiita.ac.in.
• There is no system backup for data in /home or any other partition, it is the user's responsibility to back up his/her data on a regular basis that is not stored in his/her home directory. We cannot recover any data in any of location, including files lost to system crashes or hardware failure so it is important to make copies of your important data regularly.