What mattered | Orchestration & Management | Nov 2022
Updates about GKE Gateway Controller and Azure Hybrid Benefit in AKS. CNCF advice on multi-tenant Kubernetes deployment. Container adoption report by Datadog.
News that mattered most
#1 Google Cloud’s service for managing internal and external load balancing, GKE Gateway Controller, became generally available. It lets you define routing behavior for the entire cluster with a single shared load balancer.
#2 Azure announced Hybrid Benefit for AKS, so now existing Windows Server Software Assurance and Cloud Solution Provider subscriptions can be applied on AKS.
#3 AWS added Container Build Lens to its Well-Architected Framework for building containerized applications
Most-read advice
How to solve for multi-tenancy issues – both for cluster-sharing and multi-cluster? Sub-divide tenants granularly with hierarchical namespaces, or provide each tenant with a separate virtual control plane to completely isolate their resources. A must-read CNCF blog.
Report of the month
The updated container adoption survey results from Datadog showed Nginx, Redis, and Postgre were the top technologies running containers, Istio is the leading mesh. Interestingly, it also revealed that Kubernetes updates are coming out slower than needed amidst growing container adoption.
Top Reads in Container Orchestration
#1 To help you choose between Google Kubernetes Engine and Cloud Run, use this flowchart. It incorporates deployment environments, stateful/stateless nature of services, degree of decomposition, type of architecture, and resource requirements into the decision process.
#2 To build more resilient and performant applications in Anthos or GKE, implement container-aware load balancing. This is achieved by bringing visibility at the Pods level to the load balancer using Node IP Tables or Network Endpoint Groups, and by optimizing data path latency.
#3 The cloud portability argument doesn’t stand up. In an experiment conducted by McKinsey experts, they found that migrating an application from a CSP’s managed Kubernetes to another CSP’s takes the same effort as moving it to a CSP’s native container orchestration service. Ergo, to mitigate lock-in, don’t take on Kubernetes management overheads.
#4 Now you can declaratively define and manage EMR on EKS resources like virtual clusters and job runs – as Kubernetes custom resources. This is possible through the ACK controller for Amazon EMR on EKS.
#5 How to use Vertical Pod Autoscaler in AKS. See how to enable/disable, test its installation, and automatically set resource requests on pods with it. Supports up to 500 objects per cluster.
#6 How to cut your GKE costs and your carbon footprint? Optimize CPU and memory usage by activating the ‘optimize-utilization’ profile in the GKE cluster autoscaler, routinely optimize and rightsize clusters, shut down idle or near-idle clusters, or cut costs by up to 45% by moving to GKE Autopilot.
Top Reads in Workflow Orchestration
#1 With Google Cloud Workflows, here are some best practices for serverless orchestration. Pick the right microservice communication pattern upfront and combining them when needed. Also using connectors or parallelizing when possible, avoiding URL hardcoding + ways to mitigate parsing problems, boilerplate code.
#2 Azure Logic Apps Designer to be rebuilt on open source foundations. Moving from React to Redux will enable the low-code platform to work without UI stutters; improved navigation and component search capabilities with more updates planned before the new version enters GA.