AWS User Group Edinburgh #14 - Running Production K8s on AWS Spot Instances
Description changed:
Reliable/Cheap: Pick Any Two? - Running Production K8s on AWS Spot Instances
In this talk, Guy Templeton will discuss Skyscanner's experiences of running Kubernetes production workloads on spot instances at scale. This has brought them savings of over 70% on compute cost whilst running more than 10,000 pods and serving more than 3 million requests per minute across multiple clusters. He will reveal the benefits this brings to the company, the challenges for cluster operators, and areas for improvement in how the Kubernetes ecosystem handles these challenges.
Reliable/Cheap: Pick Any Two? - Running Production K8s on AWS Spot Instances
In this talk, Guy Templeton will discuss Skyscanner's experiences of running Kubernetes production workloads on spot instances at scale. This has brought them savings of over 70% on compute cost whilst running more than 10,000 pods and serving more than 3 million requests per minute across multiple clusters. He will reveal the benefits this brings to the company, the challenges for cluster operators, and areas for improvement in how the Kubernetes ecosystem handles these challenges.
Combating Fraud and Providing Customer Insights In Real Time - Matt Houghton, CDL
Learn how CDL, a software provider for the retail insurance sector, processes vast amounts of consumer data in milliseconds using machine learning, AI and big data in order to help the UK's biggest insurance providers combat fraud, learn consumer habits, and more.
The talk will cover how we tried and failed to build a solution, called Hummingbird, with ‘big iron’ before moving to the cloud with AWS. We developed and delivered at scale in under 8 months. We will demo how Hummingbird works in production and how our DevOps team uses key AWS Services, logging, monitoring and machine learning to continually improve and deliver enhancements to the service.
We will share our experience and lessons learnt from completing an AWS Well-Architected Review and applying Chaos Engineering disciplines through Game Days. Technologies discussed include packer, terraform, Elasticsearch, DynamoDB, lambdas and containers.
Matt is a Technical Consultant at CDL. Their customer base comprises high volume retail operations or start-ups with serious growth ambitions, including over 100 Internet brands and all the major price comparison sites, and they also have partnerships with over 50 UK insurers. https://www.cdl.co.uk/
---
Agenda
6:30pm: Doors open, check-in, refreshments and networking
7pm: Reliable/Cheap: Pick Any Two? - Running Production K8s on AWS Spot Instances
7:50pm: Short break
8:00pm: Combating Fraud and Providing Customer Insights In Real Time
9pm: Wrap-up
