Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Cloud + Cloud Native [clear filter]
Tuesday, October 27

16:25 GMT

Lightning Talk: A Brisk Introduction and Demo of KEDA - Kubernetes Event Driven Autoscaling - Sagar Joshi, Microsoft
KEDA (Kubernetes Event Driven Autoscaling) an opensource framework developed by Microsoft and Redhat which enables Kubernetes workloads (Pods) to scale according to events. The built-in HPA (Horizontal Pod Auoscaler) of Kubernetes scales according to CPU/memory of pods but autoscaling with custom events like queue length or number of events or other metrics requires large amount of coding and deep knowledge of operators. KEDA framework supports autoscaling pods with a variety of event-driven systems like Kafka topics, Redis cache, Prometheus metrics and many Cloud-based service-based messaging service. In this session we will see architecture of KEDA and a demo of POD autoscaling through one of Cloud-based messaging services.

KEDA can be paired with Virtual Kubelet to implement Serverless scaling scenarios,

avatar for Sagar Joshi

Sagar Joshi

Partner Technology Strategist, Microsoft
Sagar works as a Partner Technology Strategist with Microsoft. He has a decade-long experience working with developers, architects and is a certified cloud professional with special interests in AI, ML and open-source technologies. He tweets at @sagarjms

Tuesday October 27, 2020 16:25 - 16:35 GMT
Cloud Theater

16:35 GMT

Lightning Talk: From PaaS to FaaS : Managing Serverless on Kubernetes - Suman Chakraborty, SAP Labs, India
Call it “serverless,” call it “event-driven compute” or call it “functions as a service (FaaS),” the idea is the same, allocate memory dynamically to event-driven functions to built microservices. Serverless computing platform allow developers to focus in building the application logic while abstracting away the infrastructure requirements and management details.

Many serverless offering from cloud-managed providers restrict the behavior of app logic that they run, sometimes making certain classes of applications impractical. Kubernetes has been the de-facto open source management platform for running applications in containers through a well-orchestrated tooling mechanism, taking care of the scalability, rolling updates, self-healing and load-balancing processes.
In this context, Suman Chakraborty will explain the major advantages Kubernetes provides in running serverless frameworks, the major open source projects that brings serverless functionality in Kubernetes and the challenges that has been reported by consumers in adopting Serverless frameworks in recent times.

avatar for Suman Chakraborty

Suman Chakraborty

Senior Devops Engineer, SAP Labs, India
Suman Chakraborty is a Senior DevOps Engineer at SAP Labs, Bangalore (India). He is managing and supporting DevOps for SAP ABAP on Cloud Foundry & Kubernetes platform.Suman works both as an individual contributor role as well as drives the DevOps team that builds automation framework... Read More →

Tuesday October 27, 2020 16:35 - 16:45 GMT
Cloud Theater
Wednesday, October 28

16:15 GMT

Fast Execution for Function Compositions in Serverless Computing - Ruichuan Chen & Istemi Ekin Akkus, Nokia Bell Labs
Serverless computing has rapidly gained traction with its promise of continuous scaling and low costs without the hassle of managing servers. Though it was originally aimed for running single functions, developers have embraced it to compose scalable applications with multiple interacting functions. However, many platforms map individual functions of an application into their own containers and require them to be stateless. This creates the biggest downside of serverless computing: serverless applications suffer high performance penalties from long invocation delays among functions and slow access to externalized application state. In this presentation, Ruichuan and Ekin revisit these design choices and propose an alternative with KNIX: grouping an application’s functions in the same container as separate processes while isolating different applications in different containers. Coupled with process forking and locality mechanisms, this design allows function executions to start and interact about an order of magnitude faster compared to current serverless platforms. In addition, it examines the use of stateful, addressable function executions for fast access to application state.


Ruichuan Chen

Research Scientist, Nokia Bell Labs
Ruichuan Chen received the Ph.D. degree in computer science from Peking University in 2009. He was a Postdoctoral Researcher at the Max Planck Institute for Software Systems, Germany. He is currently a Research Scientist at Nokia Bell Labs, Stuttgart, Germany. His works have been... Read More →

Istemi Ekin Akkus

Research Scientist, Nokia Bell Labs
Istemi Ekin Akkus received the Ph.D. degree from Max Planck Institute for Software Systems, Germany, in cooperation with the Technical University of Kaiserslautern, Germany. In the past, he was involved in Web privacy, data recovery for Web applications, and peer-to-peer systems... Read More →

Wednesday October 28, 2020 16:15 - 17:05 GMT
Cloud Theater

Twitter Feed