This is the first in a series of posts about controlling public cloud costs.
Public clouds are an important part of your hybrid cloud strategy – they’re powerful, scalable and highly available. They can also be very expensive: it’s possible to get a surprise bill at the end of the month that’s several times more than you planned for. One of the most common concerns people talk to me about is how to track and control public cloud spending.
I’m going to show you how the vCommander cloud management platform helps our customers minimize public cloud costs. The Embotics engineering team uses these strategies in our internal vCommander deployment to keep costs low.
vCommander provides a number of automated cost-saving strategies:
- Rightsizing – Match the instance type to the workload
- Power Schedules – Power on instances only when they’re needed
- Expiry Policy – Decommission instances when they’re no longer needed
- Compliance Policy –Tag each instance with enough metadata to indicate why it exists and which user or organization is using it
- Service Request Automation – Provision new instances through vCommander to give the admin oversight of the process and help ensure compliance
- Reporting – Get an overview of your current public cloud footprint and use trends to predict where you’ll be in the near future
So, let’s start with rightsizing. I’ll speak about this in context of the AWS public cloud, but the advice is equally applicable to any public cloud you manage with vCommander. I’d also recommend following most of these best practices for your private cloud – wasted resources are a problem, even if you’re not getting invoiced for them at the end of the month.
Picking the right instance type is vital for cost savings. Each instance needs to be just powerful enough for the workload it’s running. Any bigger and you’re wasting money – a c4.xlarge costs almost twice as much as a c4.large.
Starting with vCommander 5.5, CPU-based rightsizing recommendations are generated for AWS instances.
We gather performance metrics from CloudWatch and determine the average and peak resource consumption of the instance over a configurable period. If these remain below a threshold, vCommander recommends a change to a smaller, more appropriate instance type within the same instance family. You can configure the thresholds that trigger a change, as well as the length of the period to analyze before recommending a change.
With the configurable rightsizing introduced in vCommander 5.6, you can enforce more aggressive rightsizing thresholds for less critical workloads, leading to even more savings. For example, a you might configure vCommander to downsize a less critical workload when CPU usage is less than 50%, but downsize a critical workload only when CPU usage is less than 25%.
You can find out more about rightsizing in our post Enabling Your Users to Rightsize Their VMs.
Stay tuned for further posts on how vCommander automation enables you to control your public cloud costs.