Intel’s Snap Points the Way to Composable Monitoring
Intel continues to invest deeply in systems monitoring, judging from presentations at the Intel Developer Forum 2016 last week, where it demonstrated an integration of its open source Snap telemetry platform with Grafana, the popular data visualization, and analytics service.
Scale is creating the demand for new enterprise tooling to better understand and manage ever disparate workloads across devices, cars, buildings, etc. There were examples of this trend in almost all of the IDF keynotes and sessions we attended as well as additional discussions that focused on how Snap, Grafana, and platforms such as Kubernetes serve as ways to manage workloads and lots of data.
— alexander hazard williams iv (@alexwilliams) August 16, 2016
Snap is an open source framework for collecting, processing and publishing data via a single API from disparate metrics for a data center. It can be fed into orchestration engines or data analytics pipelines and other environments. It’s part of Intel’s “Intelligent Resource Orchestration” (IRO) approach for users to better understand and manage workloads across physical, virtualized and container-based workloads.
An important distinction to realize with Snap is that it has a pluggable architecture, decoupled in a manner that provides the user with a way to take a modular approach, something more like the Voltron animated warrior than anything else. That’s at least according to Jason Dixon, who in a blog post several years ago discussed why this cartoon superhero is symbolic of open source monitoring in our age of complex workloads and endless data streams.
— The New Stack (@thenewstack) August 15, 2016
An open source monitoring platform should be like Voltron; Dixon wrote and later presented. It has well-defined responsibilities, interfaces, and protocols. It does not require root access or an operations person to deploy. It is resilient and can withstand outages. It is automated and can model relationships between services. And it is beautiful and easy to use.
Coming full circle, Dixon now works at raintank, the company overseeing the Grafana open source project.
According to GitHub and the web pages for the project, Snap runs an agent that connects to plugins that fall into three categories: collector, processors and publishers:
- Collectors: Telemetry data gets pulled into the collectors and is designed to work with other telemetry environments such as collectd, Facter and Ohai. It also works with Intel plugins and plugins from the community.
- Processors: These plugins allow for the telemetry to be changed in some manner. The payload can be encrypted; and remove queries can be injected into workflows for tokens, filtering or other external calls, according to the Snap GitHub page.
- Publishers: The publisher plugins sink the data into another system that consumes the telemetry, according to a post by Nick Weaver when the platform was introduced in December. RabbitMQ, Kafka, MySQL, or InfluxDB are some of the platforms that work with Snap. It can also integrate with email, files, or do custom publishing to a private API.
Grafana is a popular open source platform for visualizing measurement data. It has plugins for data sources, visualization and application plugins, including ones for Snap. Like Snap, Grafana is a pluggable, composable system, one that integrates with multiple different platforms such as Graphite, Elasticsearch, Amazon CloudWatch and Prometheus. The value comes in the graphing capability, the styling, and themes that can be produced. It has authentication built in and can be annotated with data.
Snap and Grafana have a fit as an architectural pattern in the new retooling of the enterprise. They allow for a more granular way to manage complex services. In this context, we notice a new way of thinking about how to view automation and what people do well and what systems do for them.
Intel is a sponsor of The New Stack.