iSight-Cloud is a cloud-based service that allows for instant visualization of analytic query results on large datasets. Powered by the SnappyData Synopsis Data Engine (SDE), users interact with iSight-Cloud to populate the synopsis engine with the right data sets and accelerate SQL queries by using the engine to provide latency bounded responses to large complex aggregate queries.

iSight-Cloud uses Apache Zeppelin as the front end notebook to display results and allows users to build powerful notebooks representing key elements of their business in a matter of minutes.

The service provides a web URL that spins up a cluster instance on AWS or users can download the iSight-Cloud EC2 script to configure a custom sized cluster, to create and render powerful visualizations of their big data sets with the click of a button. With iSight-Cloud, you can speed up the process of understanding what your data is telling you, and move on to the task of organizing your business around those insights rapidly.

In this document, the features provided by SnappyData for analyzing your data is described. It also provides details for deploying a SnappyData Cloud cluster on AWS using either the CloudFormation service or by using the EC2 scripts.

Refer to the examples and guidelines provided in this document to help you create notebooks using which, you can execute SQL queries or data frame API to analyze your data.

Key Components

This section provides a brief description of the key terms used in this document.

  • Amazon Web Services (AWS): Amazon Web Services (AWS) is a comprehensive, evolving cloud computing platform that offers a suite of cloud-computing services. The services provided by this platform that is important for SnappyData are Amazon Elastic Compute Cloud (EC2) and Amazon Simple Storage Service (S3).

  • SnappyData Cluster: A database cluster which has three main components - Locator, Server and Lead

  • Apache Zeppelin: Apache Zeppelin is a web-based notebook that enables interactive data analytics. It allows you to make data-driven, interactive and collaborative documents with SQL queries or directly use the Spark API to process data.

  • Interpreters: A software module which is loaded into Apache Zeppelin upon startup. Interpreters allow various third party products including SnappyData to interact with Apache Zeppelin. The SnappyData interpreter gives users the ability to execute SQL queries or use the data frame API to visualize data.

Quick Start Steps

To understand the product follow these easy steps that can get you started quickly:

  1. Setting up SnappyData Cloud Cluster

  2. Using Apache Zeppelin

  3. Loading Data from AWS S3

  4. Monitoring SnappyData Cloud Cluster

Setting Up SnappyData Cloud Cluster

This section discusses the steps required for setting up and deploying SnappyData Cloud Cluster on AWS using the iSight CloudBuilder and using a script.

Deploying SnappyData Cloud Cluster with iSight CloudBuilder

Watch the following video to learn how easy it is to use iSight CloudBuilder, which generates a SnappyData Cloud Cluster.



Before you begin:

  • Ensure that you have an existing AWS account with required permissions to launch EC2 resources with CloudFormation

  • Sign into the AWS console using your AWS account-specific URL. This ensures that the account-specific URL is stored as a cookie in the browser, which then redirects you to the appropriate AWS URL for subsequent logins.

  • Create an EC2 Key Pair in the region where you want to launch the SnappyData Cloud cluster

SnappyData uses the AWS CloudFormation feature to automatically install, configure and start a SnappyData Cloud cluster. In this release, the configuration supports launching the cluster on a single EC2 instance.

It is recommended that you select an instance type with higher processing power and more memory for this cluster, as it would be running four processes (locator, lead, a data server and an Apache Zeppelin server) on it.

This method is recommended as the fastest way to deploy SnappyData. All you need is an existing AWS account and login credentials to get started!

Configuring and Launching the SnappyData Cloud Cluster

Launch the iSight CloudBuilder from http://www.snappydata.io/cloudbuilder.

  1. Enter the name for your cluster. Each cluster is identified by its unique name. The names and details of the members are automatically derived from the provided cluster name.

  2. Enter a name of an existing EC2 KeyPair. This enables SSH access to the cluster. Refer to the Amazon documentation for more information on generating your own EC2 Key Pair.

  3. Select an instance based on the capacity that you require.

  4. Enter the size of the EBS storage volume to be attached to the Amazon EC2 instance in the EBS Volume Size(gigabytes) field.


    Currently only Amazon Elastic Block Storage (EBS) is supported.

  5. Enter your email address.

  6. Click Generate.

  7. On the next page, select the AWS region, and then click Launch Cluster to launch your single-node cluster.


    • The region you select must match the EC2 Key Pair you created.

    • If you are not already logged into AWS, you are redirected to the AWS sign-in page.


  8. On the Select Template page, the URL for the Amazon S3 template is pre-populated. Click Next to continue.

  9. On the Specify Details page:

    • Stack name: You can change the stack name.


    The stack name must contain only letters, numbers, dashes and should start with an alpha character.

    • VPCID: From the drop-down list, select the Virtual Private Cloud (VPC) ID that is set as default. Your instances will be launched within this VPC.

    Click Next to continue.


  10. Specify the tags (key-value pairs) for resources in your stack or leave the field empty and click Next.

  11. On the Review page, verify the details and click Create to create a stack.


    This operation may take a few minutes to complete.

  12. The next page lists the existing stacks. Click Refresh to view the updated list and the status of the stack creation. When the cluster has started, the status of the stack changes to CREATE_COMPLETE.

  13. Click on the Outputs tab, to view the links (URL) required for launching Apache Zeppelin, which provides web-based notebooks for data exploration.
    Public IP


    If the status of the stack displays ROLLBACK_IN_PROGRESS or DELETE_COMPLETE, the stack creation may have failed. Some common problems that might have caused the failure are:

    • Insufficient Permissions: Verify that you have the required permissions for creating a stack (and other AWS resources) on AWS.

    • Invalid Keypair: Verify that the EC2 key pair exists in the region you selected in the iSight CloudBuilder creation steps.

    • Limit Exceeded: Verify that you have not exceeded your resource limit. For example, if you exceed the allocated limit of Amazon EC2 instances, the resource creation fails and an error is reported.*


To stop incurring charges for the instance, you can either terminate the instance or delete the stack. You cannot connect to or restart an instance after you have terminated it.

For more information, refer to the Apache Zeppelin section or refer to the Apache Zeppelin documentation.

Deploying SnappyData Cloud Cluster on AWS using Scripts

SnappyData provides a script that allows you to launch and manage SnappyData clusters on Amazon Elastic Compute Cloud (EC2).

Download the script from the latest SnappyData Release page. The package is available in compressed files (snappydata-ec2-<version>.tar.gz). Extract the contents to a location on your computer.


Before you begin, do the following:

  • Ensure that you have an existing AWS account with required permissions to launch EC2 resources.

  • EC2 Key Pair is created in the region where you want to launch the SnappyData cluster.

  • Using the AWS Secret Access Key and the Access Key ID, set the two environment variables, AWS_SECRET_ACCESS_KEY and AWS_ACCESS_KEY_ID.

    If you already have set up the AWS Command Line Interface on your local machine, the script automatically detects and uses the credentials from the AWS credentials file. You can find this information from the AWS IAM console.

    For example:

    export AWS_SECRET_ACCESS_KEY=abcD12efGH34ijkL56mnoP78qrsT910uvwXYZ1112 export AWS_ACCESS_KEY_ID=A1B2C3D4E5F6G7H8I9J10

  • Ensure Python v 2.7 or later is installed on your local computer.

Launching SnappyData Cluster

In the command prompt, go to the directory where the snappydata-ec2-<version>.tar.gz is extracted or to the aws/ec2 directory where the snappy-cloud-tools repository is cloned locally.

Enter the command in the following format:

./snappy-ec2 -k <your-key-name> -i <your-keyfile-path> <action> <your-cluster-name>

Here, <your-key-name> refers to the EC2 Key Pair, <your-keyfile-path> refers to the path to the key file and <action> refers to the action to be performed (for example, launch, start, stop).

By default, the script starts one instance of a locator, lead, and server each. The script identifies each cluster by the unique cluster name (you provided) and internally ties members (locators, leads, and stores/servers) of the cluster with EC2 security groups.

The names and details of the members are automatically derived from the provided cluster name.

For example, if you launch a cluster named my-cluster, the locator is available in security group named my-cluster-locator and the store/server are available in my-cluster-store.

When running the script you can also specify properties like the number of stores and region. For example, using the following command, you can start a SnappyData cluster named snappydata-cluster with 2 stores (or servers) in the N. California (us-west-1) region on AWS. It also starts an Apache Zeppelin server on the instance where lead is running.

The examples below assume that you have the key file (my-ec2-key.pem) in your home directory for EC2 Key Pair named 'my-ec2-key'.

./snappy-ec2 -k my-ec2-key -i ~/my-ec2-key.pem --stores=2 --with-zeppelin=embedded --region=us-west-1 launch snappydata-cluster 

To start Apache Zeppelin on a separate instance, use --with-zeppelin=non-embedded.

For a comprehensive list of command options, simply run ./snappy-ec2 in the command prompt.

Loading Data from AWS S3

SnappyData provides you with predefined buckets which contain datasets. When data is loaded, the table reads from the files available at the specified external location (AWS S3).


  • The Amazon S3 buckets and files are private by default. Ensure that you set the permissions required to make the data publicly accessible. Please refer to the documentation provided by Amazon S3 for detailed information on creating a bucket, adding files and setting required permissions.

  • You can also find AWS related information on the AWS homepage, from the Account > Security Credentials > Access Credentials option.

  • Information related to the Bucket Name and Folder Location can be found on the AWS S3 site.

  • If the Secret Access Key contains a slash, it causes the Spark job to fail. When you create the Secret Access Key, ensure that it only contains alpha-numeric characters.

  • When reading or writing CSV/Parquet to and from S3, the ConnectionPoolTimeoutException error may be reported. To avoid this error, in the Spark context, set the value of the fs.s3a.connection.maximum property to a number greater than the possible number of partitions.
    For example, snc.sparkContext.hadoopConfiguration.set("fs.s3a.connection.maximum", "1000")

  • To access data from AWS S3, copy the aws-java-sdk-<version_number> and hadoop-aws-<version_number> files (available in the Maven repository), to the jars directory (snappydata-<version_number>-bin/jars) in the SnappyData home directory.

To define a table that references the data in AWS S3, create a paragraph in the following format:

DROP TABLE IF EXISTS <table_name> ;
CREATE EXTERNAL TABLE <table_name> USING parquet OPTIONS(path 's3a://<AWS_ACCESS_KEY_ID>:<AWS_SECRET_ACCESS_KEY>@<bucket_Name>/<folder_name>');

The values are:

Property Description/Value
<table_name> The name of the table
<AWS_SECRET_ACCESS_KEY>:<AWS_ACCESS_KEY_ID> Security credentials used to authenticate and authorize calls that you make to AWS.
<bucket_Name> The name of the bucket where the folder is located. Default value: zeppelindemo
<folder_name> The folder name where the data is stored. Default value: nytaxifaredata

Using Apache Zeppelin

Apache Zeppelin provides web-based notebooks for data exploration. A notebook consists of one or more paragraphs, and each paragraph consists of a section each for code and results. Launch Apache Zeppelin from the web browser by accessing the host and port associated with your Apache Zeppelin server. For example, http://<zeppelin_host>:<port_number>. The welcome page which lists existing notebooks is displayed.
SnappyData provides predefined notebooks which are displayed on the home page after you have logged into Apache Zeppelin. For more information, see Using Predefined Notebooks.

Refer to the Apache Zeppelin documentation, for more information.

Functioning of the SnappyData Interpreter and SnappyData Cluster

  1. When you execute a paragraph in the Apache Zeppelin UI, the query is sent to the Apache Zeppelin Server.

  2. The query is then received by the SnappyData Interpreter which is running on the Lead node in the cluster.

  3. When the query has completed execution, the results are sent from the SnappyData Interpreter (which is running on the Lead node) to the Apache Zeppelin server.

  4. Finally, the results are displayed in the Zeppelin UI.

Connecting the SnappyData Interpreter to the SnappyData cluster is represented in the below figure.


Using the Interpreter

SnappyData Interpreter group consists of the interpreters %snappydata.spark and %snappydata.sql. To use an interpreter, add the associated interpreter directive with the format, %<Interpreter_name> at the beginning of a paragraph in your note. In a paragraph, use one of the interpreters, and then enter required commands.


  • The SnappyData Interpreter provides a basic auto-completion functionality. Press (Ctrl+.) on the keyboard to view a list of suggestions.

  • It is recommended that you use the SQL interpreter to run queries on the SnappyData cluster, as an out of memory error may be reported with running the Scala interpreter.

  • Each paragraph has its own SnappyData context. When you set a property on one paragraph, the property is applicable only to that paragraph and not to other paragraphs in the notebook.

SQL Interpreter

The %snappydata.sql code specifies the default SQL interpreter. This interpreter is used to execute SQL queries on SnappyData cluster.

Multi-Line Statements

Multi-line statements ,as well as multiple statements on the same line, are also supported as long as they are separated by a semicolon. However, only the result of the last query is displayed.

SnappyData provides a list of connection-specific SQL properties that can be applied to the paragraph that is executed.

In the following example, spark.sql.shuffle.partitions allows you to specify the number of partitions to be used for this query:

set spark.sql.shuffle.partitions=6; 
select medallion,avg(trip_distance) as avgTripDist from nyctaxi group by medallion order by medallion desc limit 100 with error

SnappyData Directives in Apache Zeppelin

You can execute approximate queries on SnappyData cluster by using the %sql show-instant-results-first directive. In this case, the query is first executed on the sample table and the approximate result is displayed, after which the query is run on the base table. Once the query is complete, the approximate result is replaced with the actual result.

In the following example, you can see that the query is first executed on the sample table, and the time required to execute the query is displayed. At the same time, the query is executed on the base table, and the total time required to execute the query on the base table is displayed.

%sql show-instant-results-first
select avg(trip_time_in_secs/60) tripTime, hour(pickup_datetime), count(*) howManyTrips, absolute_error(tripTime) from nyctaxi where pickup_latitude < 40.767588 and pickup_latitude > 40.749775 and pickup_longitude > -74.001632 and  pickup_longitude < -73.974595 and dropoff_latitude > 40.716800 and  dropoff_latitude <  40.717776 and dropoff_longitude >  -74.017682 and dropoff_longitude < -74.000945 group by hour(pickup_datetime);



This directive works only for the SQL interpreter and an error may be displayed for the Scala interpreter.

Scala Interpreter

The %snappydata.spark code specifies the default Scala interpreter. This interpreter is used to write Scala code in the paragraph. SnappyContext is injected in this interpreter and can be accessed using variable snc.

Using Predefined Notebooks

SnappyData provides you predefined notebooks NYCTAXI Analytics and Airline Data Analytics which contains definitions that are stored in a single file.

When you launch Apache Zeppelin in the browser, the welcome page displays the existing notebooks. Open a notebook and run any of the paragraphs to analyze data and view the result.

Creating Notebooks - Try it Yourself!

  1. Log on to Apache Zeppelin, create a notebook and insert a new paragraph.

  2. Use %snappydata.spark for SnappyData interpreter or use %snappydata.sql for SQL interpreter.

  3. Download a dataset you want to use and create tables as mentioned below

Examples of Queries and Results

This section provides you with examples you can use in a paragraph.

  • In this example, you can create tables using an external Dataset from AWS S3.


  • In this example, you can execute a query on a base table using the SQL interpreter. It returns the number of rides per week.


  • In this example, you can execute a query on a sample table using the SQL interpreter. It returns the number of rides per week


  • In this example, you are processing data using the SnappyData Scala interpreter. Example

  • Apache Zeppelin allows you to dynamically create input fields. To create a text input field, use ${fieldname}. In this example, the input forms are, ${taxiin=60} or taxiout > ${taxiout=60} Dynamic Form

Monitoring the SnappyData Cloud Cluster

You can monitor the SnappyData cluster using the Apache Spark Console. This enables you to observe and record the performance and the activities on the SnappyData cluster.

The Apache Spark Console displays useful information about SnappyData. This includes a list of scheduler stages and tasks, summary of tables and memory usage.

Accessing the Console

To access the Apache Spark console from the Apache Zeppelin notebook:

  1. Click on the Spark UI link provided in the paragraph. Launch Spark
  2. Once you have logged in, you can start monitoring the SnappyData cluster.

The Technology Powering iSight Cloud

iSight Cloud uses the SnappyData Synopsis Engine to deliver blazing fast responses to queries that have long processing times. Analytic queries typically aim to provide aggregate information and involve full table or partial table scans. The cost of these queries is directly proportional to the amount of data that needs to be scanned. Analytics queries also often involve distributed joins of a dimension table with one or more fact tables. The cost of pruning these queries down to the final result is directly proportional to the size of the data involved. Distributed joins involve lots of data movement making such queries extremely expensive in traditional systems that process the entire data set.

The Synopsis Data Engine offers a breakthrough solution to these problems by building out stratified samples of the most common columns used in queries, as well as other probabilistic data structures like count-min-sketch, bloom filters etc. The use of these structures, along with extensions to the querying engine allow users to get almost-perfect answers to complex queries in a fraction of the time it used to take to answer these queries.

For more information on SDE and sampling techniques used by SnappyData, refer to the SDE documentation.