Combining Out-of-Band Monitoring with AI and Big Data for Datacenter Automation in OpenPOWER

Featuring OpenPOWER Academic Member: The University of Bologna

By Ganesan Narayanasamy, senior technical computing solution and client care manager, IBM

OpenPOWER hosted its 3rd OpenPOWER Academic Discussion Group Workshop, gathering academic members of the OpenPOWER community. These members were able to share their research and developments.

One of the presenters was Professor Andrea Bartolini of The University of Bologna. The focus of his presentation was datacenter automation. Bartolini shared how this process can be implemented, examples of applications and future of work within the Power architecture.

Datacenter automation is an emerging trend that was developed to help with the increased complexity of supercomputers. To get this type of automation, heterogonous sensors are placed in an environment to collect and transmit data, which are then extracted and interpreted using big data and artificial intelligence. These technologies allow for anomaly detections, which can improve the overall learning and performance of datacenters. After information is interpreted, learned feedback is then sent back to sensors which optimizes the device.

Bartolini identifies a few specific usages that this automation process can be applied to:

  • Verify and clarify node performance
  • Detect security hazards
  • Predictive maintenance

Bartolini then focused the rest of his presentation on sharing different applications, including:

  • D.A.V.I.D.E, a supercomputer designed and developed by E4, was ranked in the Top500. This system is used for measuring, monitoring and collecting data. D.A.V.I.D.E was designed in collaboration with Bartolini and the University of Bologna.
  • Out-of-Band Monitoring: monitoring using nodes that allows for real-time frequency analysis on power supply.

Future works of this emerging practice of automating datacenters include:

  • Extending the approach of in-house security and house-keeping tasks in datacenters
  • Leveraging OpenBMC and custom firmware to deploy as part of BMC
  • Applying process to larger Power9 systems

If you’d like to learn more, Bartolini’s full session and slides are below.