Velocity: The Future is Machine Learning. Probably…

Machine Learning in Ops — Do we need it? by Hanna Foxwell, Server Density

As part of Opsview’s mission to help our customers monitor and maintain complex distributed systems, I attended the excellent O’Reilly Velocity conference held in Edgeware Road, London last week. Here’s a list of the excellent talks I attended:

Cloud native: Security threat or opportunity?

Speaker: Liz Rice (Aqua Security) — Talk details

Your organization wants to go cloud native, but you don’t want to hit the headlines as the victim of the latest hacking scandal. Liz Rice addressed the questions you need answers to: Will your deployments be less secure or more? How do DevOps processes like CI/CD and cluster orchestration affect your security profile? And what can we all do to minimize the risk of exploits?

Why an (interactive) picture is worth a thousand numbers

Speaker: Miriah Meyer (University of Utah) — Talk details

Feeling overwhelmed by huge amounts of data has become the norm. Creating, effective visual representations of data offloads some of the work of quickly finding interesting patterns to our powerful perceptual system. Miriah Meyer explored the role that interactive visualizations can play in helping us find meaning in mounds of data and discusses the limitations of this approach. For example, Opsview Monitor ships with a thoughtful set of dashboard elements and predefined reports to help address this growing need.

Scaling a startup with a 21st century language

Speaker: Christopher Meiklejohn (Instituto Superior Técnico) — Talk details

Christopher Meiklejohn is building an application that helps users select a bottle of wine based on the wines that they enjoy, using a new programming language called Martinelli. Christopher offered an overview of Martinelli, highlighting the key features of this new language that allow the fault-tolerant, highly scalable operation of his application.

Quantifying scalability with the Universal Scalability Law

Speaker: Baron Schwartz (VividCortex) — Talk details

Distributed systems used to be the exception, but today they’re the norm; so it’s more useful than ever to be able to quantify scalability. Baron Schwartz explained how to use the Universal Scalability Law to characterize how your systems truly behave, why they may not scale as expected, and how to improve them.

Machine learning in ops: Do I need it?

Speaker: Hannah Foxwell (Server Density) — Talk details

Machine learning is the new big data. Everyone is supposed to be on board, but do we understand why? As platforms become more complex and change more frequently than ever before, Hannah Foxwell argued that it’s time we stopped trying to maintain them manually. During Hannah’s excellent talk she explored the technology and real use cases for machine learning in infrastructure operations and SRE.

My takeways

On the Tube home I reflected that monitoring is still a core function of DevOps but that legacy platforms struggle to add value to modern, dynamic IT operations. It’s exciting to see how many academics and monitoring vendors are working together on innovative projects to make business sense of the the millions of data points they collect every day. The industry’s probable future appears to be greater use of machine learning and the thoughtful data visualisation leveraging the best practices established in innovative industries like bio-tech.

Here at Opsview, we’re evolving with the times, as well. We currently have automatic monitoring of systems via Autodiscovery with Detection Mapping. We’re working on additional, simplified approaches, that handle the complex back-end data for you, while still automatically monitoring your environment. Stay tuned.