Approaching Cloud Optimization!

In this article, I briefly want to talk about what's been my experience and what are the key pointers to this exercise.

a. Continuous Exercise - Optimization does not happen as a one-time activity, needs continuous evaluation or observation- we have to keep the customer's buying and spending trends in watch, application(s) hunger to consume the resources, unauthorized/un-monitored accesses to the accounts, unused resources check and this can take time.

b. Using only the required resources- With the vast offerings by AWS its always misleading to get carried away with the term 'latest or what's new' as human beings we will always keep looking for what's new and may procure resources which may not be required for your applications (please note I am not advocating that don't use the latest but use the optimal resources). The concept of 'Right Sizing' is coined by AWS itself to be able to help a customer use the optimal and the right resources- is your customer aware of this ?.

c. Look out for these?-

Data Transfer- The cost of transferring data into your cloud is usually free. But transferring it out is another matter. So, in same way you should design single-cloud systems to minimize the cost of internal data transfers, you should also aim for a cost-efficient multi-cloud architecture.

Be mindful that outbound network charges vary from vendor to vendor. And also that, at individual vendor level, the cost of data transfer between different cloud regions varies from geographical region to region. There are lot of other factors to deep dive, I am leaving those for further but definitely every customer environment will be different.

And, finally, don’t forget you can reduce the network load on your cloud servers by offloading web traffic to a content delivery network (CDN). Depending on your application requirements, this could be a third-party solution, such as Cloudflare, or in-house cloud offering of AWS - CloudFront.

Compute and Storage- AWS offers a range of different instance types and storage services aimed at different use cases, cost requirements and performance expectations. So finding the best fit for your workloads can be a complex challenge—all the more so when comparing resources across several clouds, ideal way would be to focus on fitment in the beginning itself.

Discount alternatives- Discounted alternatives to standard virtual machine pricing offer the potential to significantly reduce the running costs of your cloud environment. They typically come in the form of billing constructs, such as AWS Reserved Instances, and disposable capacity- AWS Spot Instances.

Whats new to help you ? (as on today- 19 Aug 2019)

AWS launched on 24 Jul 19 a new feature called “EC2 Resource Optimization Recommendations,” which does exactly what the name promises. It’s not flashy, it’s not especially exciting, but it may just save you and your company a good amount of money (and maybe that’ll get you that raise you’ve been hoping for).

The resource optimization tool will look at your EC2 usage and give you personalized recommendations to find idle and underutilized instances. To do this, it looks at your usage history, CloudWatch metrics and your existing reservations. When it finds an idle instance, that is, one that has lower than 1% maximum CPU utilization, the tool will recommend that you shut it down. No surprises there. When it finds underutilized instances, it’ll present you with three different sizes that you can move to that’ll likely fit your usage patterns better than your current plan.

One caveat: this feature currently works for all standard EC2 instances, but it’s not available for GPU-based instances yet.

This new feature is now available to all AWS users. You can find it in the AWS Cost Management suite, where it’ll join the rest of AWS’ tools for keeping an eye on your budget and how you’re spending it.

There are a lot of external tools available in the market which will provide additional help for this exercise.

That's all folks from my end. The above article is a culmination of data from various resources which I read and can provide the same if asked.

Thank you

Shashank Roy