Team Opsani is the place where we pull back the Opsani curtain and share stories about our team and our internal goings-on. This month, we sat down for a chat with Chris Li, our principal engineer.

Hey Chris! First things first: How does being a principal engineer differ from simply being an engineer?

Being a principal engineer requires more leadership and depth.

You have an impressive resumé on you. How do you think your experiences with the Linux kernel and VMWare empower your work at Opsani?

The Linux kernel experience is definitely useful in the sense that most of the cloud service is using linux. Both Linux and VMware give me a chance to work on very complex systems with high performance. A lot of those experiences will be able to transfer into the cloud. The container is a Linux environment.

Tell us about your main project, the Opsani Learning Predictive Autoscaler project. What is it and how do you expect it to improve the overall Opsani product?

The ALOS project is basically a smarter way to Kubernetes HPA. In this webinar, I talk about current limits of the Kubernetes HPA, and why it is hard to use for customers. It brings down to 1) HPA by default using CPU as an indication to scale out. That is not effective for non CPU limited service. Other metrics can be set up but require more setup and limited documents. 2) The behavior of the HPA is very sensitive to the HPA target metrics value. It is not easy for customers to pick the perfect value.  3) After the scale up decision, the newly spawn up pods will take some time to spin up and serve the request effectively. The time delayed control is a typical hard problem for control theory.

OLAS will address those problems by learning the HPA pod behavior and figure out the right time and right moment to scale up. Both saving customer’s money and preserving the service level objective (SLO).

Where else do you think the Opsani AI could branch out? 

Many places. The current cloud space has huge spending and Opsani proves that there are a lot of places to make things better by introducing automation and making them smarter. Anything that requires repeat human labor to maintain can potentially be automated.

Let’s talk about some of your other areas of expertise. What’s your current impression of the blockchain? How should people immersed in cloud technologies be thinking about the blockchain?

How many hours do you have?! I think there is definitely value in blockchain technology. There is also definitely hype and bubble in the current blockchain movement as well. I would think it is hard to find those niche markets where blockchain can really shine to solve some real problems and provide real value. As you can tell, I don’t advertise “everything with blockchain is so much better.”

How about flight control systems for drones, another area of interest for you. Does the AI used here have any parallels with the Opsani AI?

Actually, it does. A big part of the flight control is real time control systems. OLAS is a control system, it is actually much easier than flight control. In flight control we need to process the sensor data and generate control output every 1ms. In HPA it is 1 minute at most. Also the cloud behavior is much more consistent than the flight control system, which is subject to wind, camera lighting conditions etc.