Matei Zaharia on New Developments in Apache Spark

What was the original vision of Spark? How did it hold up? Matei Zaharia speaks to this, and new developments like DataFrames and Spark SQL, on November 2nd at IBM Research, Yorktown NY.

“The Spark project started at UC Berkeley to provide a general engine for distributed data processing, capable of combining many types of analytics into complex workflows. It was one of the first projects to offer a functional API for big data processing, and it has grown to include the largest integrated standard library for big data, with support for relational queries, streaming, machine learning and graph processing. We discuss lessons learned bringing Spark to developers and building out its library.

In particular, although Spark’s functional API led to concise code, we found that it could limit opportunities for optimization, both in CPU time and memory usage. We have developed a new API called DataFrames that gives significantly more information about data and computations to the engine, through an interface based on records with a known schema. A further extension, Datasets, offers a typed API on top of DataFrames that integrates into Java and Scala in the same way the original Spark API did. Finally, more and more of Spark’s standard libraries are written to take DataFrames / Datasets as input, enabling rich optimizations such as loop fusion and join reordering across libraries. We describe ongoing research at MIT to use this new interface for optimizations. Together, these changes are one of the first attempts go beyond the functional APIs proposed for big data processing while maintaining ease of programming and composabilty.” – Matei Zaharia

Matei Zaharia is an assistant professor of computer science at MIT and CTO of Databricks, the company commercializing Apache Spark. He started the Spark project during his PhD at UC Berkeley. He is broadly interested in large-scale computer systems and networks, and has also contributed to projects including Mesos, Hadoop, Tachyon and Shark.

Things Matei is too modest to tell you: ACM gave Matei the best doctoral dissertation award for 2015, and he received two Best Paper awards at NSDI 2012 and SIGCOMM 2012.


You Might Also Enjoy

Gidon Gershinsky
Gidon Gershinsky
18 days ago

How Alluxio is Accelerating Apache Spark Workloads

Alluxio is fast virtual storage for Big Data. Formerly known as Tachyon, it’s an open-source memory-centric virtual distributed storage system (yes, all that!), offering data access at memory speed and persistence to a reliable storage. This technology accelerates analytic workloads in certain scenarios, but doesn’t offer any performance benefits in other scenarios. The purpose of this blog is to... Read More

James Spyker
James Spyker
3 months ago

Streaming Transformations as Alternatives to ETL

The strategy of extracting, transforming and then loading data (ETL) to create a version of your data optimized for analytics has been around since the 1970s and its challenges are well understood. The time it takes to run an ETL job is dependent on the total data volume so that the time and resource costs rise as an enterprise’s data volume grows. The requirement for analytics databases to be mo... Read More