Jim Scott

Subscribe to Jim Scott: eMailAlertsEmail Alerts
Get Jim Scott via: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn

Blog Post

Taking Apache Spark for a Spin | @BigDataExpo #BigData

What is Spark?

You might have looked at some of the articles on Apache Spark on the Web and wondered if you could try it out for yourself. While Spark and Hadoop are designed for clusters, you might think you need to have lots of nodes.

If you wanted to see what you could do with Spark, you could set up a home lab with a few servers from Ebay. But there’s no rule saying that you need more than one machine just to learn Spark. Today’s multi-core processors are like having a cluster already on your desk. Even better, with a laptop, you can pick up your cluster and take it with you. Try doing that with your rack-mount servers.

What is Spark?
If you’re looking to try out Apache Spark, it helps to know what it actually is. Spark is a cluster computing framework that builds on Hadoop to support not only cluster computing, but also real-time cluster computing.

Spark consists of the Spark Core, which handles the actual dispatching, scheduling and I/O. Spark’s key feature is the Resilient Distributed Dataset, or RDD. RDDs are the basic data abstraction, containing a distributed list of elements. You can perform actions on RDDs, which return values, and transformations, which return new RDDs. It’s similar to functional programming, where functions return outputs and don’t have any side effects.

Spark is so fast because it represents RDDs in memory—and because RDDs are lazily evaluated. Transformations will not be calculated until an action on the RDDs has been requested to produce some form of output.

Spark also gives you access to some powerful tools like the real-time Spark Streaming engine for streaming analytics and the MLlib machine learning library.

Installing Spark

Installing Spark is easy enough. While the MapR distribution is essential for production use, you can install Spark from the project website on your own machine, whether you’re running Windows or Linux. It’s a good idea to set up a virtual machine for exploring Spark, just to keep it separate and reduce the possible security risk of running a server on your machine. This way, you can just turn it off when you don’t need it. Linux is a good choice because that’s what most servers will be running.

You can install also install Spark from your favorite distribution’s package manager. At least with the package manager, you won’t have to worry about dependencies like Scala. You can also install them from the respective websites or even build from source if you want.

Using the REPL

One of Spark’s greatest strengths is its interactive capabilities. Like most modern languages, Spark offers a REPL: A Read-Eval-Print-Loop. It’s just like the shell, or a Python interactive prompt.

Spark is actually implemented in Scala and you can use Scala or Python interactively. Learning both of these languages is beyond the scope of this article, but Python tends to be more familiar to people than Scala. In any case, if you’re interested at all in technologies like Spark, you likely have experience in some programming, and either Scala or Python shouldn’t be too hard to pick up. Of course if you have experience in Java that will work as well.

When you’ve got Spark up and running, you’ll be able to try out all the actions and transformations on your data.

The Spark equivalent of a “Hello, world!” seems to be a word count.

Here is an example shown in Python:

text_file = spark.textFile("hdfs://...")


text_file.flatMap(lambda line: line.split())

.map(lambda word: (word, 1))

.reduceByKey(lambda a, b: a+b)

You can see that even in Python, Spark makes uses of functional programming concepts such as maps and lambdas. The Spark documentation has an extensive reference of commands for both Python and Scala. The shell lets you quickly and easily experiment with data. Give it a try for yourself to see what Spark can really do.


If you’ve been curious about Spark and its ability to offer both batch and stream processing, and want to try it out, there’s no need to feel left out just because you don’t have your own cluster. Whether you’re a developer, a student, or a manager, you can get a taste of what Apache Spark has to offer. When you’re ready for production use, opt for the MapR Spark distribution when you’re ready for a complete, reliable version.

To further explore Spark, jump over to Getting Started with Apache Spark: From Inception to Production, a free interactive ebook by James A. Scott.

More Stories By Jim Scott

Jim has held positions running Operations, Engineering, Architecture and QA teams in the Consumer Packaged Goods, Digital Advertising, Digital Mapping, Chemical and Pharmaceutical industries. Jim has built systems that handle more than 50 billion transactions per day and his work with high-throughput computing at Dow Chemical was a precursor to more standardized big data concepts like Hadoop.