Berlin CloudNativeCon + KubeCon 2017
Last week my colleagues and I had the opportunity to attend the 2017 CloudNativeCon + KubeCon in Berlin. The journey was quite enlightening. Although we listened to several talks about the future of Cloud Native technologies, for me, the most interesting part was that the present of Kubernetes already looks quite bright. From the numerous talks, I wanted to share the two case studies that I think represents this the most.
One of the case studies presented the now-infamous shut-down of Amazon’s S3 server. The shutdown affected countless services that rely on its advertised high availability. The speaker, with some satisfaction, announced, that due to the automated management capabilities of Kubernetes, companies that used the fledgling open-source service remained available through the incident. They simply watched Kubernetes switch hosts while the rest of the internet was melting down in panic.
Another case study of interest was that of Magazine Luiza, a Brazilian retail company, that recently adopted Kubernetes. This adoption was particularly useful during Black Friday, when traffic to online stores increases exponentially. The company has a history of shaving off more and more downtime during the traffic rush on Black Friday. However, in previous years they were unable to produce 100% availability, which is crucial, as on Black Friday every second of downtime means significant revenue losses. In the preparation to the 2016 event, they decided to give Kubernetes a try. While deploying the apps they even developed an open source deployment tool. Mr Pereira, the representative of Luiza Labs, speaking at the event, proudly announced that in 2016 Black Friday went smoothly, without any interruptions in service.
As these case studies show, Kubernetes as a technology, although still in its early stages, is already making strides in transforming the limits of highly available services. It exceeds the human restrictions that prevented swift reactions to spikes in traffic or unexpected resource-outages. The automated deployment and scaling it provides seems to have significant potential, and we have seen already what it can do.