It’s a time of more: more data, more workloads, and more powerful technologies. So it’s critical to ask ourselves how cloud native can help improve our environmental footprint. Here’s what we heard at today’s keynotes.
Keynotes: Revolutionize cloud native architectures
The cloud native landscape is evolving rapidly and using more resources as it expands. Taylor Dolezal, head of ecosystem, Cloud Native Computing Foundation, asked end users to present to tell their stories of sustainable computing, beginning with a look at WebAssembly. Michelle Dhanani, principal software engineer, Fermyon made the analogy that cloud computing is like boxing. Virtual machines are the heavyweight runtime of the cloud – powerful, but slow. Containers are the middleweight with a balance of speed and power. WebAssembly is the “lightweight” third wave, as it’s nimble and works in milliseconds. And now with SpinKube, an open source stack, it’s possible to streamline the experience of developing and deploying WebAssembly workloads on Kubernetes and leverage the power faster and more sustainably.
And if you want to only use one stack? SpinKube has been submitted to CNCF and will give you ability to install, run, and optimize with a single stack
Dolezal invited Kai Walter, distinguished architect with the ZEISS Group, to describe how the company learned to do more with less post–deployment when asked to operate within some tight parameters including same or higher amounts of data, fewer resources, lower costs and leveraging existing skills and environments. Using Spin with WebAssembly, ZEISS Group was able to reduce density from 400MB to 2MB and cut costs by 60% with a 43% performance gain.
Ralph Squillace, principal product manager with Microsoft, described how open source foundations make innovations possible including:
- Kubernetes: Its maturity and stability allows us to innovate around it.
- The ContainerD project allows contributors to make improvements without affecting containers already running
- RunWasi allows workloads to run in the same pod as containers.
- KWASM is used to install spin-shim easily.
Keynote: Awards Ceremony
The CNCF community is built of three main pillars: the Governing Board, the Technical Oversight Community, and End Users. Alolita Sharma, AIML observability engineering at Apple announced the newly formed End User Technical Advisory Board (TAB). And there were awards recognizing exceptional end users in the ecosystem. Third and 2nd place went to Expedia and Shopify.
This year’s winner is CERN! In his acceptance speech, Ricardo Rocha from CERN described how the shift to cloud native has been a game changer. He acknowledged all the contributors, maintainers, and end users who make OS possible.
Keynote: Building IT Green
Gualter Barbas Baptista, lead consultant for platform strategy and enablement at Deutsche Bahn, spoke about Building IT Green. Deutsche Bahn is trying to make train transport for freight and passengers more sustainable and it cannot be done with infrastructure. The company must digitalize, but that requires massive amounts of energy. An initiative to address this question was started, with a grassroots push from developers to help. But how to know if they were being successful? Their cloud-based efforts needed cloud-based tools to measure the footprint, but none existed. So they had to figure out for themselves how to measure efficiency and consistency.
The company had a three-tiered approach:
- Technology Platforms: Build them with standard optimization for efficient cloud operations with Kubernetes as the base. They added scheduling with kube-downscaler to manage and turn off workloads as needed.
- Visibility: Kepler and Grafana allowed them to evaluate energy statistics at a high-level to container-level to understand consumption at the single component level. They did need to update for security and enterprise readiness and they were able to contribute to make this update
- Data: Data is so important and dashboards are beautiful. Monitor everything!
Deutsche Bahn is still in the early stages of knowing how to measure energy. Barbas Baptista recommends companies in a similar situation should start small and focus on empowering developers to make the best decisions.
Keynote: Innovate responsibly
Aparna Subramanian, director of production engineering at Shopify, tackled an important topic: how to get the most out of existing infrastructure while scaling efficiently and sustainably, during his keynote “Innovating Responsibly: How to Navigate Sustainability in the Era of Kubernetes.” Shopify has massively grown their e-commerce infrastructure and has been thinking about how end users and platform operators approach this, what sustainability looks like, as well as the need for it to be a shared responsibility.
Subramanian introduced Adrienne Jan from Scaleway, a full-stack cloud provider. At Scaleway, the biggest source of carbon is in the data center where keeping the servers cool is a high resource activity. Cloud providers must manage power and water efficiently to become more sustainable. Scaleway has had success significantly decreasing usage by recommending cloud users run something they call “code sobriety audits.” The audits address the issue of complex code using too much power.
David Meder-Marouelli, 1&1 Mail & Media GmbH, and Todd Ekenstam of Intuit to discuss their approaches to sustainability. According to Meder-Marouelli, the biggest sources of inefficiencies are servers. Every server running Kubernetes is using wattage on idle. Large well-utilized machines are the answer and he provided an example of how to better optimize the number of servers required. He recommended measuring what exists and then writing an automation to provision and deprovision the resources needing optimized.
Ekenstam spoke on the power of autoscaling to handle the heaviest loads and reduce resources when no longer needed. It’s key to size and scale workloads just like is done with servers. Extra resources for “peace of mind” are expensive and not sustainable. Kubernetes provides ways to automatically and dynamically scale workloads, but configuring these programs can be complicated. To solve this at Intuit they are using AI to build a way to intelligently autoscale.
Subramanian agreed that Compute is a great starting point and ended the talk with a few more efficiency ideas to consider:
- Databases: how many database replicas are too many?
- Do storage buckers have a lifecycle management policy?
- Do bot traffic and user traffic have the same latency requirements?
- How many autoscaling workarounds can we undo?
- Are bigger clusters more efficient than smaller clusters?
Let’s work together to make cloud sustainability actionable!
Also, check out the recent announcement that CNCF has partnered with Udemy! This new partnership will endorse content to support CNCF’s audience of more than 233,000 project contributors and millions of current and aspiring cloud native developers across the globe to help them prepare for various CNCF certification exams.