machine learning as a service

New Major Release V11 Sept. 24th 2020: ProActive Workflows & Scheduling (PWS) and Machine Learning Open Studio (MLOS)

Try new version online or on premises

ProActive Workflows & Scheduling 11.0 is out and available for trial and download (free trial version for limited infrastructure downloadable from your account). This new version includes a lot of features that enhance this already robust and flexible workload automation solution, recognized by Gartner in its “Market Guide for Service Orchestration and Automation Platforms”.

Enjoy the following major features

  • Subscribe to notifications on job status updates or on Job Planner issues. Choose how you receive notifications: through the portal, email or third-party interfaces.
  • Access Cloud Automation services based on roles and deploy through a secured proxy.
  • Display job executions in a Gantt chart with Job Gantt portal to better visualize job chaining.
  • Visualize resource consumption in a gantt chart using Node Gantt portal and easily optimize resource usage.
  • Enjoy ProActive portals in your preferred language: English, French, Portuguese or Spanish.
  • Manage files that will be used by the workflows with the new Dataspaces graphical interface.
  • Benefit from new Job Planner configurations to better control over the submission behavior.
Try now

New in Machine Learning Open Studio

The Machine Learning Open Studio part, dedicated to machine learning professionals and data scientists, has been updated with the following:

  • New tutorials and documentation
  • New workflow templates for Horovod, TensorFlow, Keras and PyTorch
  • Support to multiple containers (docker, podman, singularity) on AI workflows
  • Machine Learning as a Service (MLaaS) with Data Drift Detection
  • JupyterLab, COCO Annotator and OpenRefine available as a Service
  • Support to fetch satellite images from Copernicus
Try now

Full release notes

  • ProActive portals are now available in English, French, Portuguese and Spanish
  • Users can now upload their company logo on the ProActive portals
  • A modal is now available to manage graphically the user space from the Scheduler and the Automation Dashboard or when submitting workflow with the USER_FILE variable type
  • A modal is now available to manage graphically the credentials when submitting a workflow with the CREDENTIAL variable type

Scheduler

  • Forked and Non Forked mode available at the task level
  • New variable types: NOT_EMPTY_STRING, HIDDEN, CREDENTIAL, USER_FILE, GLOBAL_FILE, USER_FOLDER, GLOBAL_FOLDER
  • New variable model syntax to define variables as optional
  • Submission window to display the variables in the same order as they are written in the workflow
  • New script engines available for VBScript and PHP
  • Resource Manager API accessible from workflow tasks
  • XML Workflow can now be retrieved and opened in the Studio
  • New filter to sort jobs by submission, start, and finish time

Resource Manager

  • Available policies are automatically suggested depending on the selected infrastructure
  • Current user credentials now used by default
  • Advanced parameters are hidden by default to ease configuration, and can be visualized and modified by selecting Advanced configuration
  • Elastic infrastructure is now labelled
  • A link to documentation is available for each infrastructure and policy to help users with configuration
  • Display the cost of the Azure Infrastructure in the node source
  • Allow node reconnection after PAMR router restart
  • Node tokens can be managed from the portal

Studio

  • Cut, copy and paste of tasks with Ctrl+X, Ctrl+C and Ctrl+V
  • Change task variables at submission time

Catalog

  • Ability to update a bucket group after creation

Workflow Automation

  • Retrieve job output from Workflow Automation portal
  • Improve results visualization with the “Result List” column and Result Map in the Past Jobs section

Cloud Automation

  • Run it on multiple remote infrastructures
  • Fault tolerance and recovery mechanism
  • Access deployed services through a secured proxy
  • Role-based access of deployed services

Job Planner

  • Redesigned calendar creation view shows when the calendar is saved and updated
  • Multiple time zones support
  • Chaining jobs are now displayed in Job Planner Gantt
  • Improve association visualization by modifying the icon according to association status (PLANNED, DEACTIVATED or ERROR)
  • Export and Import associations
  • New configuration to submit jobs on missed executions
  • New configuration to deactivate submission by putting association in the error status when the submitted workflow fails
  • New configuration to postpone the execution when a submitted job is ongoing
  • New configuration to synchronize Job Planner housekeeping with the Scheduler housekeeping

Analytics

  • A new Job Gantt view to visualize workflow executions in a Gantt chart
  • A new Node Gantt view to visualize resource consumption in a Gantt chart
  • New filters (before and after finished time, variable names, results) and new operators (not equal, does not contain, starts with, less than, greater than)
  • Specific visualization for jobs with issues

Notifications

  • Subscribe to notifications
  • Configure the notification method per subscription (portal, email or third-party)
  • Notifications on job / task status change (submitted, running, finished, in-error, paused, resumed)
  • Support for Job Planner notifications
  • Scheduler notifications supported on status changed or if the database goes down
  • New filters available per notification severity and per notification type

New workflows

  • Task template to send results by email

ProActive Jupyter Kernel

  • New pragmas to display job results
  • New pragmas to handle if, loop and replicate workflow mechanisms

New tutorials

  • Web portals walkthrough
  • Workflow in-depth
  • APIs walkthrough
  • Scheduler Administration
  • Create a deep learning workflow from scratch
  • Hyperparameter optimization
  • Data visualization using Visdom and Tensorboard
  • Machine Learning as a Service
  • Hands on Proactive Jupyter Kernel
Request demo

More news...

Published on October 14, 2020