Welcome note by the organiser and the sponsor

Efficiency of software components became the new norm. Hypervisor is a vital component of any virtual platform and Qemu-KVM is the most popular. Managed by libvirt and more high-level orchestration tools like OpenStack, it offers many options for tuning. In this talk we will look at tweaks that hypervisor offers, will go through conceptions of NUMA and CPU affinity on a provider's side; as well as some options that is available on the user's side. From the perspective of hardware we shall see how optimization affects computation efficiency using AI-models as a juicy example

Kubernetes as many of us know provides a declarative interface for the infrastructure & state we desire. Traditionally, in order to have efficient recovery (should anything catastrophic occurs), we save the yaml files in a folder, link to a script file and probably forget about it. As our project grows, files gets more and more complex, and potentially out of sync too. Our confidence that the yaml files are of the current state of the cluster will deteriorate too. Hence GitOps with Flux. By using nearly the same declarations and methodologies that we are familiar and already apply on our cluster anyways, we place git as the source of truth for our cluster state. Any changes to the cluster must go through git. This enables clear accountability and traceability, while enabling contributions from those that don’t have access to the cluster. This session also will look into how Kwong's company got into the GitOps bandwagon little by little, with little to no impact to production workloads