Cloud Cost Optimization: 6 fundamental Steps

Cloud Cost Optimization: 6 fundamental Steps

Gartner forecasted that the cloud service market will grow at the same pace as
it is growing from the last few years and by 2022 90% of the organization will
be using cloud service. The reason behind this growth is the benefits cloud
offers to the organizations. Less workload hosting cost is one of its benefits.
But when we start using cloud services offered by different public cloud
players (i.e. AWS, Azure, and Google) we realize that we are paying even more
then what we were paying On-premises and there are many reasons behind
this which we shall discuss later in this blog. Cost blowout is among the top
challenges of businesses. Sometimes we even get surprises at the end of the
month when we receive the surprise bill that is several times more than we
planned for.

Let’s discuss these steps which we can take right away to avoid surprise bill
at the end of the month.

1. Rightsizing

Rightsizing does not only optimize cloud cost but also help in achieving the
peak performance from the resources we are paying for. If I talk about the
manual approach which I used to follow two years back was to analyse
performance matrix of each VMs or EC2 instances every two-three months and
if I find compute resources are underutilized (Let’s say less than 50%). I used to
bring down the VMs configuration. But this is a time taking process and

sometimes prone to error and finding the correct VMs size is also a tedious

The second way is to use CMP (Cloud Management Platform) tools. To take
care of this. It saves our lots of time and effort both. These tools keep track of
the VMs performance and give the recommendation of the correct VMs size
and potential cost benefits almost on an everyday basis. I have been personally
using Kumolus to take care of cost optimization and automation part in my
AWS environment.

2. Power Automation

There are many types of workloads which do not need to be running 24X7. For
example, our dev VMs do not need to be running all the time, as developers
mostly use dev resources between Monday-Friday and during the day time
only. We can schedule auto shutdown on weekends as well as after 7 pm every
day and start them again next morning at 7 am. It can allow saving up to 65%
on computing cost.

3. Opt for RI

This is again a good option provided by almost all the public cloud providers. If
you find that your compute resource requirements are going to be same for
the next one to three years. I will suggest, invest in RI. Cloud providers gives
discount up to 75% on RI. It’s a good saving option to opt for.

4. Remove Unused/Orphaned Resources

There were many times when I provisioned VMs for tasting or for some other
purposes and forgot to delete (Even after deleting the VMs some of the
attached resources don’t deleted and it remains there in our cloud account) or
stop it. With me, it generally happens when I am stuck in multiple tasks and at
the end of the month, it comes me as a surprise cloud bill. If the cloud
environment is small these unused resources can be tracked with minimal
effort. But if the environment is huge it becomes cumbersome to manage. To
avoid such events from happening I prefer CMP tool (Kumolus).

5. Idle resource

Next step which we can do to optimize cloud cost is to identify idle resources.
In case of idle resource compute resource CPU utilization is between 1-5%,
however, the organization is paying 100% of the resource cost. We can
leverage cloud features like auto-scaling, load balancing, and on-demand
provisioning. These cloud offers can help in scaling our computing at any time.

6. Get bulk credit from the cloud provider

When you are hosting or planning to host a considerably bigger environment
and paying big some of the amounts to your cloud provider. You can ask them
to give volume discounts.

There is no doubt that the cloud holds great potential. But it will be cost-
effective only when we are paying close attention daily. In my opinion, CMP is
a good and cost-effective option to keep track of the above steps.