"Organisations are spending precious and costly time gathering, configuring, testing and troubleshooting machine learning environments for their data scientists, at the expense of time that could be used to deliver insights to the business", stated Bill Wagner, CEO of Bright Computing. "This new product eliminates the drudgery of getting a machine learning environment up and running, along with Hadoop or Apache Spark, in a cluster-ready environment that can automatically scale up as your demand for machine learning capacity grows."
Last year, Bright began offering tested and verified deep learning libraries and frameworks to its customers. Since then, demand for machine learning has grown, and Bright has responded by greatly expanding its machine learning portfolio, adding more libraries, more frameworks, and deepening the integration with its management software suite. Now, with the addition of big data, Bright offers a complete set of tools allowing data scientists to accelerate their work.
Bright Cluster Manager for Data Science integrates with all of the leading deep learning frameworks including Tensorflow, NVIDIA CUDA Deep Neural Network library (cuDNN), Deep Learning GPU Training System (DIGITS), MXNet, Caffe, Caffe2, pyTorch, and CNTK as well as key big data platforms: Spark, Hadoop, and Cassandra.
The new product also includes jupyter, Zeppelin, and DIGITS notebook front ends data scientists can use to perform interactive queries without having to use the command line.
It also provides a best-in-class management solution for the hardware that lies at the heart of deep learning installations: NVIDIA and AMD GPUs, and Intel accelerators. As a result, customers can choose their numeric computation accelerators of choice.
"Our mission has always been to make managing infrastructure easy so that our customers can focus on their work", Bill Wagner stated. "This new offering enables us to extend that mission to data scientists working with compute-intensive, data-intensive, deep learning projects."
Bright Cluster Manager for Data Science, including the ability to build custom versions of included packages using EasyBuild, will be demonstrated at Supercomputing 17, November 13-16 in Denver Colorado, at booth #937.