Coral vs Azure Advisor

Use both, starting with Azure Advisor, which is free and needs nothing installed. It shows recommendations for all your subscriptions. Then add a tool where Advisor can only point at the fix.

• Use Advisor across Azure: idle VMs, unattached disks, reservations and savings plans.
• Use Coral where your spend sits in AKS or EKS clusters. Coral rebalances CPU and memory requests, works with your Horizontal Pod Autoscaler (HPA), and drains spare nodes so your autoscaler can remove them.
• Use Amoeba where VM data disks are bigger than the data on them, on Azure or AWS.

Advisor stops at Azure. On AWS, Coral covers EKS the same way: see EKS cost optimization. For the rest of your Azure spend, read our guide to Azure cost optimization.

Kubernetes places each pod by its requests, not by what it uses. The scheduler keeps the sum of requests on a node below the node’s capacity. So a node can be full while its CPU sits close to idle.

The Cluster Autoscaler on AKS counts requests too, when it works out how busy a node is. So you reduce requests first. Then the autoscaler has spare nodes to remove.

Advisor’s cost list has five recommendations for AKS:

• Run the Vertical Pod Autoscaler (VPA) in recommendation mode, for request and limit values.
• Tune the cluster autoscaler profile, so unneeded nodes go sooner.
• Turn on AKS cost analysis, to split cluster cost by Kubernetes constructs or Azure assets.
• Use Spot nodes for work that can handle interruptions.
• Move to Prometheus-based Container Insights, to reduce monitoring cost.

The first two go after requests and spare nodes, and both leave the work with you. Run that way, the VPA calculates values but changes no pod. So someone sets them, workload by workload.

Microsoft also doesn’t recommend the VPA beside an HPA that scales on CPU or memory. HPA vs VPA explains why the two clash.

Coral does that work, and keeps doing it. It sizes requests to real use and works with your HPA. Coral also empties spare nodes, so the Cluster Autoscaler or NAP removes them.

Our Kubernetes cost optimization guide walks through both steps by hand, and AKS cost optimization shows what Coral sets up on AKS.

Advisor’s disk advice has three parts:

• Review disks that aren’t attached to a VM (in preview).
• Move busy Standard HDDs to Standard SSD.
• Consider reservations for managed disks.

None of its cost recommendations makes a disk in use smaller. Amoeba covers that gap.

Amoeba works on the data disks of running VMs: Azure managed disks and Amazon EBS volumes, on Windows and Linux VMs. It moves a VM’s data disks onto one btrfs pool sized to your data, with room to grow.

The pool grows before it fills. When your data goes down and stays down, Amoeba shrinks it one disk at a time.

• Delete what you don’t need first: find unattached disks in Azure.
• Price a disk with our Azure disk pricing calculator.
• See what Amoeba does on Azure: Azure managed disk cost optimization.

Microsoft Learn, “Introduction to Azure Advisor” (opens in a new tab); Microsoft Learn, “Cost recommendations - Azure Advisor” (opens in a new tab); Microsoft Learn, “Optimize virtual machine (VM) or virtual machine scale set (VMSS) spend by resizing or shutting down underutilized instances” (opens in a new tab); Microsoft Azure, “Azure Advisor” (opens in a new tab); Microsoft Learn, “Vertical Pod Autoscaling in Azure Kubernetes Service (AKS)” (opens in a new tab); Microsoft Learn, “Cluster autoscaling in Azure Kubernetes Service (AKS) overview” (opens in a new tab); Kubernetes, “Resource Management for Pods and Containers” (opens in a new tab).