Hello everyone, my name is Y. You know, last time I talked about the most popular Kubernetes cost management tool, and this time I want to talk about one of the coolest Kubernetes optimization tools from More Exel Data. As you can see from the screen, I'm a multi-cloud guy. I love Kubernetes, and feel free to follow me by any one of the links. And why I'm covering Exos Data first. So, this morning when I woke up, I saw quite a lot of LinkedIn posts about Exos. So, basically, the company has been featured by Gartner in the latest 2024 Gartner Hypercycle for Platform Engineering and the Site Reliability Engineering under the Autonomous Workload Optimization. So, that checked my interest to record a video about Exos Data first. If you haven't seen the Hypercycle, here is what it looks like, and the category mentioned here is about Autonomous Workload Optimization. Okay, so who is Exos Data? So, in short, it's a startup company. They pioneered the cloud optimization.

company founded in 2018 by con alumina, and two of the key points I want to highlight is live migration of sta for workload without an interruption and a plus to ensure you can run your workload highly reliable from the spot instances. Okay, what it looks like so the product name called X infrastructure Optimizer. So we are to using Amazon eks as an example. So because Amazon e is so popular, so as you can see from the animation, you can live migrate one application or work cloud from one instance to another instance, or maybe you want to move back. The beauty here is your state for application continue to run. There is no interruption. All your Services remains uninterrupted, and your network settings, your network IP address, your connections, your memory, your K driver, Etc still remain the same. So that's what they call nondisrupted level migration. Okay, why we do this? Yeah, if you migrate from on Dem instance to spot instance, all variously, it can help you save a lot, lot of money. So behind of the scene how it works, let's

see you have an E cluster. You have a control plane here. You got an E node running as a work node, and you might be running Carpenter already. Yeah, Carpenter is the most popular open-source cost optimization tool, and it can allow you to easily schedule your workload across all different instance types rather than the native auto-scaling group. It's a little bit challenging. Yeah, you need to create the node group for different instance types. So, nowadays, companies are more popular. I do see a lot of the E EKS customers are using Carpenter. So, if you're already using Carpenter, you don't have to replace anything. You continue to use Carpenter for scheduling of your workloads across all different instance types. So, what's going to be added here with the E X O S T A X I O here is it allows you to create what they call nested VM or virtual E node. So, it can be running from on-demand instances or spot instances. But if you want to save money, apparently, you might choose to run from a spot instance. And let's say you already

running on a spot instance and the instance for the spot instance. There will be an interruption, so just before it's going to be interrupted. With the excess status feature functionality, you can live migrate your whole workload to another EC2 instance. Again, you can use a spot instance or EC2 instance. If you use a spot instance, you can save lots of money. So once you spin up the new instance, you can live migrate all of your workload to the new instance. So literally just live migrate the whole nestm or virtual e node. Okay, and once the migration is successful, it will terminate the existing instance. So that's how EXO St of the xio for E can help you to save money, being able to run the application or workloads highly reliant on the spot instance. Remember, there is no interruption while the migration is happening. So one of the typical use cases, so it will be a perfect use case for your database application. So could be running PostgreSQL, MongoDB, etc., or could

be you're running any data pipeline, so you want to run your machine learning, your training, or your inference. It takes a long time. You don't want to restart from the beginning, so without the interruption, being able to run on the spot instance, even if there is an interruption before the interruption, EXO still allows you to migrate the workload. But think about its lift-shift from one E2 instance to another one, and then there is no instruction to your workflow. And the other two use cases are for your M Ops or your CI/CD pipeline. So if you're running your CI/CD pipeline, it might take a long time. If there is any interruption of the spot instance, it will cause you some problems. You might take a long time to reset up the whole environment. So with these capabilities, you continue to run your CI/CD pipeline and being able to run all these workloads from the spot instance. That's where excess data can help you to save a lot of money. So let's see a demo, but not

enough okay. I haven't built the love demo environment yet, so once I build up the love demo environment, the one thing which I really want to show you is how you can live migrate to one of the state of application from one instance to the other and without any interruption or disruption to the service. Okay, I will record another video about the demo, so for now, let me wrap up with two of the links here. So, one is Kasuga, which is the most active Kubernetes and AI User Group. Yeah, please feel free to join us. We already have 7,000 followers. The other one is KS DM, that's the most active Kubernetes Discussion Group. If you haven't joined us, yeah, you should. There are nine sou Plus members already joined us. That's pretty much all I want to talk about for now. Thank you for watching. I hope it is useful to you. Enjoy your day.

Exostellar Intro - 1 of the coolest Kubernetes Cost Optimization tools

Scripts