Follow Datanami:
October 31, 2016

Bigstep Adds Spark Service to Bare-Metal Cloud


A Spark-based analytics engine released last week runs on a low-latency, software-defined bare metal fabric and can scale Spark clusters and the IT infrastructure that supports them.

Bigstep, which has dual headquarters in London and Chicago, said its real-time Spark service is designed to speed deployment of real-time data streaming applications as more Spark implementations shift to the cloud. Among the emerging uses for such implementations are the Internet of Things and algorithm decision-making, Bigstep Founder and CEO Lucas Roh noted in a statement.

Roh added that Bigstep’s bare metal cloud platform would help reduce development requirements for its real-time Spark service. The company also said it would offer a “pay-per-use” container-based Spark cluster optimized for real-time streaming applications that use multiple concurrent Spark contexts.

A rapid prototyping feature includes a built-in Jupyter interface for Scala, Python or R programming languages. Jupyter Notebook is designed to allow data scientists to combine code, graphs, dashboards and descriptive texts within the same document while performing operations interactively.

Meanwhile, real-time data streaming applications run in parallel with other container application clusters.

The company added that its bare-metal fabric could run multiple Spark versions, including Spark 2.0, using the same pool of resources.

The introduction of the Spark platform follows the company’s release in early October of a real-time application container service designed specifically for streaming applications. The company said it container service also targets emerging infrastructure based on micro-services and memory-intensive workloads requiring low latency and higher performance.

The container service is based on Docker and can run distributed streaming applications on the company’s bare metal cloud. Those applications can be built on Spark Streaming, Apache Flink or the Heron streaming replacement for Apache Storm, the company said.

Persistent storage requirements have recently overtaken security as the top barrier to adoption of application containers in production. Hence, Bigstep stressed its real-time container service offers high-end persistent support, including storage volumes that follow containers as they move across clusters.

Meanwhile, the company said customers could deploy its new Spark service alongside either Zoomdata and Bigstep data lakes, or use it with applications deployed on-premises or on its managed container service.

Recent items:

Unraveling Hadoop and Spark Performance Mysteries

Overcoming Spark Performance Challenges in Enterprise Hadoop Environments