Microsoft has announced the release of the Microsoft Azure Data Science Virtual Machine, or DSVM, created to run on Windows Server 2016 Datacenter Edition. Using the new version, users can manage big data analytics on cloud-based virtual machines without the need to configure, install, or maintain project hardware.

The new DSVM is currently available through the Azure marketplace, and is tailored to meet the needs of developers exploring data science, machine learning, and deep neural network applications.

The DSVM can run on either CPU-only or GPU-based virtual machines, and it is available in Ubuntu or Linux.

With the DSVM, users can have a data science desktop that is entirely cloud-based, so it is scalable for each project or task as required. Flexible configurations allow users to scale up or down as required, and environments are repeatable so VMs can be eliminated when they are no longer needed.

The new version of the DSVM is based on the latest Windows 2016 Data Center edition, and includes many new features and add-ons that were not available in previous versions of the DSVM.

These additions to the Data Science Virtual Machine include Docker container support, which allows users to create and run containers on Azure cloud networks, and the addition of Office ProPlus which will let any user with a valid license create documents in Word, Excel, OneNote and PowerPoint.

The new Windows Server 2016 DSVM also includes tools to facilitate deep learning and cognitive computing. First, Microsoft has added unified support for deep learning on GPU or CPU-only VMs by pre-installing drivers and the latest GPU versions of deep learning frameworks CNTK, TensorFlow, and MXNET.

Also, the company has upgraded the DSVM to R Server 9.1, which includes cognitive modelling and ‘enterprise-grade operationalization’ for scaling and scoring virtual machines. Pre-trained cognitive models in Server 9.1 are sentiment analysis, which assesses the sentiment of a sentence, paragraph or phrase automatically. The Image Featurizer, on the other hand, is a pre-trained cognitive model that can derive up to 5,000 separate features of an image and compare the similarity of two images on the separate feature points.

Using the DSVM, data scientists can explore data and develop local models on R Server and Python, or experiment with data on a Jupyter notebook on browsers running Python or enterprise R Server. Azure resources can be administered and scaled, and the DSVM can be used to share storage and datasets or code with all members of a team.