Share your experience using AWS Compute Optimizer

My primary use case for Spot revolves around managing and optimizing my compute infrastructure across AWS, mainly EC2 and EKS on a daily basis, such as monitoring the workloads and ensuring that Kubernetes and EC2 instances are provisioned, scaling, and replacing spot instances for ephemeral tasks.

Recently, I have been using Spot for Kubernetes through Ocean, which continuously analyzes pod capacity requirements and adjusts the node groups accordingly. It helps handle node drains, rebalancing, and spot interruption gracefully without requiring custom automation. I have been using it recently for provisioning ephemeral tasks for applications such as stateless applications.

Another main use case for Spot revolves around managing ephemeral tasks or stateless application workloads. Additionally, I focus on cost optimization and optimization of my overall work nodes so that I do not need to manage everything or the nodes for my workloads.

Spot overall is a really good tool, and the best feature is the intelligent autoscaling engine it has. It looks at pod requests, historical usage, interruption risk, and pricing trends to choose the best instance type at the lowest cost. The fact that it can mix dozens of instance families dynamically is something which the AWS native autoscalers simply cannot do. I also love the zero-downtime node replacement.

Spot's zero-downtime mode replacement has helped my team by enabling us to reduce our operational overhead. I no longer maintain custom scaling scripts, which has saved several hours or weeks of SRE and DevOps time.

Spot has impacted my organization positively by delivering a very clear ROI. My compute costs have reduced, capacity and production output have increased, and my overhead for maintaining custom scripts or doing some of the tasks manually has been saved. It has saved significant time and resources and given my organization a better competitive edge.

I think overall Spot is quite good. The UI is powerful, but it can feel a bit dense for some new users. More guided onboarding would help teams adopt advanced features faster, and deeper insights into Kubernetes resource usage would be beneficial. Overall, it is a really good platform that is quite mature and stable.

I chose a rating of nine because it may sometimes be a bit overwhelming for newcomers, and there are also a few areas in which the EKS Kubernetes level granularity is a little missing. Overall, I think Spot is a really good and stable tool.

I have been using Spot for two years.

Spot is stable and quite mature.

Spot's scalability is quite good, and it can be expanded to multiple environments.

The customer support is quite good. I did not need to reach out to them a lot, but I think it was good.

Positive

Previously, I was using native solutions only.

Before choosing Spot, I evaluated other options. I was totally dependent on native tools, but I assessed some alternatives such as Cast AI, which was a combination of a few tools including Cast AI and Carpenter for EKS. However, I chose Spot because it provided more automation, smarter instance selection, and had features of many other alternative tools combined in one.

I purchased Spot through the AWS Marketplace.

I cannot share the exact number, but I can share the figure that my EC2 and EKS compute cost reduced by around 65 to 75% with Spot. This was because I was able to reduce some workloads for stateless applications, which needed ephemeral storage for running tasks and processing. This has helped me a lot because my overall operational overhead has been reduced significantly, and I save a lot of time from otherwise managing the servers or maintaining everything properly.

I have seen a return on investment as I was able to reduce my compute cost by 60 to 65% for some of my applications that needed light temporary processing tasks with ephemeral storage and stateless applications. It has also reduced some of my overheads because I do not need custom scaling scripts. It has saved me several hours a week for my key SREs and my DevOps teams.

The experience with pricing, setup cost, and licensing has been quite good. The pricing is reasonable and convenient, and the value it offers is completely in line with what I am spending. I think this is really good. However, the exact licensing part is handled by some other members of my team.

I would definitely recommend Spot for those looking to run their workloads on AWS, especially those needing cheaper compute for tasks that will not require much for temporary processing. For scaling up and down existing applications, Spot is really one of the best investments you can make, especially for EKS. I gave this product a rating of nine out of ten.

Our main use case for Spot is to optimize our cloud cost, specifically the AWS cloud cost, by buying proper savings plans and commitments so that we can effectively manage the overall compute savings that we are doing. That is the immediate use case, and in the future, we would like to use Spot Instances and conduct a POC with that to reduce further AWS costs.

In the future, we would like to utilize Spot Instances and implement some automation in place to leverage that aspect from AWS, thereby reducing the overall running compute costs.

The best features Spot offers include the overall tool, the dashboard, and the major impact it has in our organization, which is to reduce costs from both a commitment management perspective and with Spot Instances along with the automation behind it.

The dashboard and automation features stand out to me because currently, we do not have a mechanism to detect whether a particular instance is going to expire within AWS when we are using a Spot Instance. Spot helps us detect such changes within AWS at least one hour prior, allowing us to have a mitigation plan in place and host other servers within the same setup to avoid impacting the application. This automation has helped us significantly since previously we were trying to use Spot Instances but were unable to implement that. One use case we plan to implement within the next few months is the overall commitment savings, which we used to manage manually, but with Spot, it automatically helps us detect when and what kind of instances to purchase, making it easier for us to manage our overall commitment for the savings plan.

Spot has impacted my organization positively so far, showing some cost savings to a certain extent. I believe that over the next few months, we will obtain a proper number in terms of cost savings, but I can already see certain cost savings from the past few months.

Spot could be improved by adding features that help identify data to optimize storage costs by detecting datasets within our environment based on criteria such as age and usage.

Storage optimization is the only need on my wish list for Spot.

We have been using Spot very recently, so it has been close to a year that we are using it.

Spot has been stable so far.

I have not yet checked the scalability of Spot because we are using it for all the services running on AWS. As of now, I feel it is meeting our expectations, though I have not tested it for scalability at higher volumes.

My experience with customer support has been good.

Negative

We did not previously use any other solution.

Deploying Spot in my environment was quite easy; it did not take long at all, just a few hours to integrate with AWS.

My experience with the configuration process was good; it was straightforward as we followed the documentation and were able to complete it successfully.

I found the procurement process easy since we completed it through the AWS Marketplace, making it straightforward.

There was nothing concerning regarding pricing, setup costs, or licensing, as it was processed through the Marketplace and based on the savings model.

I have not seen a return on investment yet. I believe it will take another four to five months to get the return on investment.

My experience with the metering and billing has been very good.

There is nothing specific that I did not like about the metering and billing experience.

We did not evaluate other options before choosing Spot. We were looking at other kinds of tools, but none had the same kind of features as Spot.

I purchased Spot through the AWS Marketplace.

My experience with Spot so far has shown a few noticeable changes. We are not renewing the new savings plan that we were doing earlier, and the changes we have been implementing have helped us to reduce some fraction of cost. However, I believe it will take another five to six months to reflect the actual savings since certain EC2 instances we are running still have an existing savings plan. Once those savings plans expire, we will be able to get the maximum savings out of it.

I find these two features quite useful for us and we have been using them. Spot is very new to us, and over the next few months, we may evaluate more features and possibly conduct a POC with those.

As of now, I can forecast approximately 5% to 7% cost savings that I have noticed.

Cost saving is the main benefit, and the manual effort required for reviewing and purchasing savings plans and other RIs is something we do not have to manage manually now, as we get automatic suggestions from Spot. This additional aspect has helped us significantly. It will take a few more months for us to further review the other features that this product has.

I rated Spot as a four out of five because storage optimization would make it a five for me.

My advice for others looking into using Spot is to definitely proceed with it for a more robust solution in terms of cost management. Spot helps manage savings plans and RIs that are difficult to handle manually. Leveraging the automation feature for managing Spot Instances can provide additional benefits.