Redis Accelerates Spark By Over 100 Times

Redis Labs, the home of Redis, today announced its integration with Spark SQL and the release of the Spark-Redis connector package. Redis Labs’ benchmarks using time-series data show that running Spark on Redis as a data store results in 135 times faster processing compared to Spark using HDFS and 45 times faster processing compared to Spark using Tachyon as an off-heap data store or Spark storing the data on-heap. 

The Spark-Redis connector package is open source and provides a library for writing to and reading from a Redis cluster with access to all of Redis’ data structures – String, Hash, List, Set, Sorted Set, bitmaps, hyperloglogs – from Spark as RDDs. In addition, the package also ensures close cluster alignment between Spark and Redis clusters, reducing network overhead and ensuring optimal processing times. The main advantages of using Redis with Spark include:

• Acceleration of Spark performance by over 100 times, in use cases such as spark-timeseries
• Redis data structures allow elements of data to be accessed individually and rapidly, minimizing serialization/deserialization overhead and avoiding transfer of large chunks of data.

Big data is coming of age and customers are demanding that big data insights are extracted in real-time,” said Yiftach Shoolman, co-founder and CTO of Redis Labs. “This is where Redis Labs fills the gap by delivering both the right performance and optimized distributed memory infrastructure to accelerate Spark. Our goal is to make Redis the de-facto data store for any Spark deployment.”

The Spark-Redis solution enables:

• Redis data structures exposed via Spark RDD and DataSet API
• Spark SQL support (via DataFrame and DataSource API) as a standard query interface
• Use of Redis Cluster as a distributed memory infrastructure for Spark

Additional planned enhancements to the solution include using the combination of Spark and Redis for other popular use cases such as graph computation and machine learning.

The Spark-Redis connector package was developed by Redis Labs in close collaboration with Databricks”, says Patrick Wendell, vice president of engineering at Databricks. “Spark and Redis are a powerful combination enabling sophisticated analytics with a great deal of simplicity and speed.”

Sign up for the free insideBIGDATA newsletter.