The IT focus on cost-cutting is not new. But managing your monthly cloud bill is different from trimming operating costs and wringing efficiencies from on-premises purchased hardware.
Unlike buying infrastructure for your data center, which includes a one-time outlay to a hardware vendor, buying cloud services means that someone cuts a check to each cloud vendor every month.
The pay-per-use cloud service scenario has advantages, but also creates complications. Each instance, storage, and network use may get lost in a bill that includes literally hundreds of line items. Accounts payable may write one check each month to your cloud vendor, but linking each line item back to the group that is paying for it can be tricky. It is easy to overprovision resources and end up paying for resources that don’t add value. And, the person that initially requested the service may stop using the service without deleting the instance, leaving a wasteful remnant on the cloud bill.
The goal with cost optimization in a pay-per-use environment is to simply avoid consumption that doesn’t add value. So now the question is: How exactly do you do it?
You can use analytics and automation take a bite out of your next cloud bill:
- Inventory: Take stock of all workloads, storage, and network traffic, and link back to those who are consuming the service.
- Reporting: Report back usage to the project, group, or business unit that requested the service. This can enable show-back or hard charge-back to those paying for the service.
- Optimization: Use automation to avoid paying for services that don’t add value.
Here are some simple ways to use automation to optimize your bill, after you take inventory and sort reporting.
- Right-size instances. This is the low-hanging fruit of cloud cost reduction. IT has a habit of over-provisioning “just in case.” But in a pay-per-use and auto-scaling environment, it is better to use a smaller instance size for each tier of frequently deployed workloads. Then, use deployment automation to default to the standard configuration, either through self-service or IT help request.
- Turn it off when you are done. This is another easy one. Automate the deletion of short-lived workloads after a preset period of time. Set a policy to delete workloads for developers after two weeks or after two days for QA and test stages of the lifecycle—whatever appropriate timeframe makes sense for each use. This can also be done for certain types of temporary workloads or even for anything related to an underfunded project. Sending a notification prior to deletion will ensure that users can extend the workloads if needed.
- Suspend it when you are not using it. Automate suspension and restoration of workloads that are used sporadically by specific users (e.g. developers) on a pre-set schedule, such as off at 7 p.m. and back on at 7 a.m. Additionally, give users an easy way to click an exception if they get inspired and want to work in the middle of the night.
- Scale it when you need it. Use scaling policies to automatically horizontally scale more instances as usage grows and remove instances as usage shrinks. You can trigger scaling based on simple infrastructure metrics like CPU utilization. But it is even better to use more powerful application performance management (APM) tools like AppDynamics® to instrument your entire application ecosystem. You can detect and respond to emerging user metrics or business process metrics to ensure quality of service where it counts most.
- Have a plan. If you have hard cost limits for projects, teams, or even individuals, remember that people respond well when they know the limits and can easily monitor their own usage.
- Set a budget. Let people know someone is monitoring cloud usage and costs against that budget and communicate limits in a way that matters to these individuals.
It pays to pay attention to your cloud bill. In a pay-per-use environment, use management tools and automation to avoid paying for services that don’t add business value. This could pay for your ‘free lunch’ (or even free lunches for a year).
Visit the CloudCenter Suite page for the full scope of how we can help you make multicloud management easier.
If you would like even more information on the features and functions of Cloud Center Suite, watch this on-demand webinar.
Excellent blog, Amarjeet!