Will Google DeepMind Artificial Intelligence Transform Data Center Infrastructure Management (DCIM)?

Google Data CenterWhen it comes to hyper-scale data centers and countless servers, storage and network infrastructure management who knows more about site performance and reliability than Google! But just like other hyperscale and enterprise data center owners and operators, Google is plagued by huge data center energy costs that can add to operational expenditure and thereby reduce operating margins for data center operators.

Data Center equipment such as servers and storage infrastructure devices generate huge amounts of heat and most of the energy expenditure is in trying to keep them cool to elongate life cycle and also ensure that the equipment is performing at desired levels. This cooling is typically accomplished by using large HVAC equipment such as pumps, chillers and cooling towers. In Google’s case, these servers are powering Google Search, Gmail, YouTube, etc which serve millions of users and need to be kept operating smoothly without downtime.

Data Center Infrastructure Management (DCIM) emerged as a software category focused on providing data centers the ability to monitor power and environmental conditions beyond the rack level to cover data center aisles and overall facility.  It helps data centers to implement optimum cooling and reduce downtime by offering real-time alerts. DCIM software help in managing the Power Usage Effectiveness (PUE) of data centers which is a metric used to determine the energy efficiency of a data center and is the ratio of total amount of energy used by a data center facility to the energy delivered to computing equipment – with overall efficiency improving as the PUE value decreases and gets closer to 1. DCIM offerings offer PUE management dashboards as a measure to improve energy efficiency in the data center facility. But PUE is only part of the story.

The problem with traditional DCIM software is outlined in a recent joint blog by Rich Evans, Research Engineer, DeepMind and Jim Gao, Data Centre Engineer, Google:
  1. The equipment, how we operate that equipment, and the environment interact with each other in complex, nonlinear ways. Traditional formula-based engineering and human intuition often do not capture these interactions.
  1. The system cannot adapt quickly to internal or external changes (like the weather). This is because we cannot come up with rules and heuristics for every operating scenario.
  1. Each data center has a unique architecture and environment. A custom-tuned model for one system may not be applicable to another. Therefore, a general intelligence framework is needed to understand the data centre’s interactions.

In order to solve the above challenges, researchers at Google’s DeepMind group have been applying DeepMind’s machine learning to Google data centers and have managed to reduce the amount of energy by up to 40 percent. The implications are significant for Google’s data homepage-energycenters that are already running on renewable energy, given its potential to greatly improve energy efficiency and reduce emissions overall. With DeepMind’s algorithms, Google can help deal with climate change impacting its renewable energy data centers. As per the blog, Google has invested heavily in renewable energy and has already seen significant benefits in energy savings.

 

Reducing energy usage has been a major focus for us over the past  10 years: we have built our own super-efficient servers at Google, invented more efficient ways to cool our data centres and invested heavily in green energy sources, with the goal of being powered 100 percent by renewable energy. Compared to five years ago, we now get around 3.5 times the computing power out of the same amount of energy, and we continue to make many improvements each year.

With Google’s DeepMind technology,  the energy savings and operational efficiencies will go multiple steps further. Companies adopting Google’s Infrastructure as a Service (IaaS) or Platform as a Service (PaaS) on Google Compute Engine (GCE) or Google App Engine (GAE) cloud will also also benefit and report improvement in their enterprise energy efficiency.

So What has Google Done with DeepMind’s Artificial Intelligence (AI) capability?
neuralGoogle used a system of neural networks trained on different operating scenarios and parameters within its data centers to create a more efficient and adaptive framework to understand data center dynamics and optimize efficiency. The Google-DeepMind researchers incorporated historical data from sensors for temperatures, power, pump speeds, setpoints, etc to analyze within deep neural networks. Google’s PUE results are already very positive. For this exercise, Google trained the neural networks on the average future PUE and along with other parameters they were able to predict the future temperature and pressure of the data center over the next hour. This is pretty significant and goes beyond what traditional DCIM players can offer.pue-average
Below is a graph from a live Google Data Center that shows the analysis when machine learning recommendations were off or on. When it was on, Google’s DeepMind lab saw 15 percent reduction in overall PUE which was the lowest ever value seen at that site.
DC_PUE.png
Since Google’s DeepMind algorithm is a general-purpose framework to understand complex dynamics,  Google plans to apply this to other challenges in the data center environment and beyond in the coming months. Given that many legacy industries are getting connected in the Internet of Things (IoT) world, Google may be able to use the technology to improve power plant conversion efficiency (getting more energy from the same unit of input), reducing semiconductor manufacturing energy and water usage, or helping manufacturing facilities increase throughput. The practical and commercial use of predictive analytics with Google’s DeepMind technology is definitely an area to watch closely.

Are you an expert in IoT, Big Data, Virtualization, Cloud and/or IT technologies?Do you want to feature yourself or your content on VcloudNews.com? Do you have products you would like to showcase to our visitors? Or simply do you want to share your comments with our readers? We want to hear from you and have both free and sponsorship opportunities for you and your products. E-mail us at submit@vcloudnews.com to join the fun.

Leave a Reply

Your email address will not be published. Required fields are marked *