SUSE is looking closely at the potential use-cases of machine learning within the software stack.
I sat down with Dr. Thomas Di Giacomo (also known as Dr. T), Chief Technology Officer @ SUSE to talk about the expansion of the office of CTO across the globe, what are the emerging trends that the company is focussing on, how it’s using machine learning and much more.
Talking about the office of CTO, Dr. T said that the primary goal is to increase engagement with the open source communities and bring back that feedback to the company. Another goal is to localize SUSE’s presence in some key territories like North and South America and APAC region. It will enable SUSE to have leaders with expertise in different areas in which the company operates, including cloud and containers.
Talking about emerging areas where he sees potential growth for SUSE, he said there is a lot of growth and adoption around software-defined storage. “Improvements made in projects like Ceph will help enterprise customers to take it to new use-cases,” said Dr. T.
SUSE and Machine learning
It’s true that SUSE is not building a machine learning platform. It doesn’t have a TensorFlow based product today, but the company is working on integrating TensorFlow with some of the SUSE technologies.
Dr T cited many reasons for SUSE’s interest in machine learning. One is that SUSE engineers like to work on emerging technologies like these. Simple. The second and more business-centric reason is that some of the SUSE partners and customers are working on integrating machine learning capabilities into their own solutions that are running on SLES (SUSE Linux Enterprise Server), SUSE OpenStack and SUSE Containers. “Our partners like SAP and HP have interest in those technologies and we are working with them to see how we can integrate all of that,” said Dr. T.
He cited use-cases to improve the software stack of infrastructure with machine learning. SUSE has some proof of concept internally to filter the bug requests and the support requests using machine learning.
“There are use-cases where we are using machine learning to check how much time it takes to deploy packages on, for example, 500 servers. It allows users to predict the maintenance window or how much time it will take in a different environment. So we are looking at machine learning to help operations inside our products,” he said.
SUSE is quite bullish about software-defined storage solution and it is using machine learning in that area as well. Dr. T said that if you look at software-defined storage or even SUSE CaaS Platform. The distributed nature of these products with distributed systems with clusters makes it very complex to set-up.
“As humans, we try to configure things but it’s not perfect. We think that machine learning could help,” he said. Having many customers who use this technology creates a pool of data about how it’s configured, how it performs and machine learning can take that data and help SUSE create configurations that are more optimized automatically. As a result, you get fully optimized set-up configurations that continue to get better with time.
Dr. T said that HPC is another area where machine learning can be used.
You can watch the interview on YouTube and listen to it on SoundCloud.