Transcript: Making Public Cloud Cost Possible | NetApp IT’s Spot Use Case

Leyddy Portas: Good afternoon, everyone. And welcome to this NetApp on NetApp INFORM Session. I’m going to be your host today. My name is Leyddy Portas, and I am a technical partner manager at NetApp. Just a quick note. For those of you who are new to these NetApp INFORM webinars, we are hosting them every month. So, if you are not receiving the invitations through me, just please contact me and I’ll be happy to add you on our distribution list. I’ll drop my email on the chat as well.

We have a lot of people on the call today. So, if you have any questions during the presentations and demo, you can just post them on the chat. And we’ll be responding to the questions as we go along. Today’s session is entitled NetApp IT Spot Use Case.

Presenting today, we have from NetApp Tirou Chidambaram, who is a domain architect at NetApp IT, and he has played a key roll in deploying and maintaining VMware virtualization platform in NetApp IT for over a decade. And now he’s transitioned to the cloud. He evaluated Spot, and he’s the lead on the deployment of Spot in NetApp IT. Tirou, thank you for joining us. And I’ll hand it over to you.

Tirou Chidambaram: Thanks, Leyddy. As Leyddy introduced, I am Tirou Chidambaram. I am a domain architect at NetApp. My main focus is migrating workloads from on-prem data center to cloud, and as well as cost optimization in cloud. Today, I am going to talk about Spot by NetApp, formerly known as NetApp acquired, and today we rebranded the name as Spot by NetApp. 


I will talk about Spot basics, just to level set, everyone understand about the Spot service, the Spot instance service that AWS offers, and native Spot instances. And, I will talk about the Spot by NetApp advantage. So, how Spot by NetApp comes in and makes things easier adopting Spot instances. And, at NetApp how we are adopting Spot instances, and I will also show some savings and demo. Here, the demo, actually, I am using the screenshots. In the interest of time, we are not showing the actual demo here, but I will show the screenshots from our systems.

AWS Spot Instances: What is Spot Instance?

Okay. What are Spot Instances? Spot Instance(s) are nothing but the excess capacity in AWS EC2 access capacity. So, when a customer goes and asks for EC2 instances, he gets the on-demand instances, which is always available. We have this notion in the cloud, right? Cloud capacity is infinite. So, how hyperscalers make sure they have infinite capacity? Basically, by provisioning large volume of excess capacity, so that any customer, and as for EC2 instances, there will not be any shortage of EC2 instances. When they have this large excess capacity, it’s also ideal in their data centers. Basically, AWS wanted to monetize this excess capacity. So, they came out with the Spot Instance idea.

There are normal instances, which is called on-demand instances, and this excess capacity in AWS, they call it as Spot instances. So, when you go and ask for resources from this EC2 instances, from this excess capacity, AWS gives it to you with a steep discount. Up to 70%, 80% discount you get.

AWS Spot Instances Basics: Trade Off

But what is the trade-off? The trade-off here is AWS can take back your instance at any time. They just give you a two-minute notice, and then they can take back your instances.

You are running your application, then all of the sudden AWS comes … somebody else, some other customer asked for on-demand instances, then they take your Spot Instances and give those to Spot Instances and on-demand instances to their customers requesting for the same. If you are running some applications and all of the sudden, with your applications, you lose your applications, since you lose your instances.

AWS Spot Instances Basics: How to mitigate spot terminations

So, this is where Spot by NetApp comes in.

What Spot by NetApp does is, if you provision your application by your Spot by NetApp, then your application gets provisioned on EC2 instances. The Spot by NetApp periodically snapshots your application, basically the root volume. It snapshots the root volume. And then, in (the) case (that) termination happens, Spot by NetApp goes and search in Spot capacity, then closely matching instances to the instance which just got terminated. So, using the snapshot data stack, and it re-provisions the instance on that other available Spot Instances.

Basically, this overhead, let’s say if there is no Spot by NetApp, then you provision your instance on Spot Instances. And then it gets terminated, you have to go and look for available Spot Instances, which is closely matching the terminated instance. Then, prior to that, you have to take snapshot backups, then you have to restore. And all these overheads, Spot by NetApp basically removes all this overheads and makes it automated. This re-provisioning basically takes 5 to 10 minutes, depending on your root volume size.

Spot by NetApp

Spot by NetApp, the product, basically, it’s a SaaS-based solution. You don’t need to install any agents and things like that on the EC2 instances. It’s a SaaS solution. It’s ideal for stateless applications. When I say stateless applications … you know (I mean) … containers and web servers …  any kind of stateless application, it’s ideal. Because, see, when it comes to stateful application, there is this 5 to 10 minutes downtime, this re-provisioning downtime. But when it comes to stateless application, if it is containers, even within Spot, actually, you get that two-minute advanced notice. The AWS gives that two-minute advanced notice before you terminate your instance.

These two minutes advanced notice is very well enough to drain your containers and re-provision your containers on other parts. So, that’s why the stateless, Spot is ideal for stateless workloads. But, it can also be used in stateful applications. For instance, all your sub-prod machines, it would be fine to have this 5 to 10 minutes downtime, right? When it comes to the provisioning, this 5 to 10 minutes re-provisioning time for sub-prod workloads, usually, it’s okay. So, all your stateful applications for sub-prod, you can leverage Spot. You can also, when it comes to state auto scaling groups or any other fault tolerant applications, also.

Let’s say you have web servers behind load balancers. There also you can provision your application on Spot instances, because let’s say you have your web farm. It has 10 web servers. Losing two, three web servers, it’s still okay. Actually, your web sessions, you can redirect to other web servers and your application can continue to run. So, wherever you can, you have this kind of if we can handle this five to 10 minutes re-provisioning time, wherever you can handle (it), you have architected your application to handle this re-provisioning time in all those places, whether it is sub-production or production. You can use Spot instances.

Even when it comes to production application, tier-three, tier-four application, five to ten minutes, downtime is still fine in most of the cases. So, there also, though it’s your production application, you can Use Spot instances, if you can tolerate that 5 to 10 minutes downtime once in a while. So, the Spot termination, basically, it doesn’t happen every day. It can happen randomly. There, it can happen once in a week. Or even for a couple of months, your instances may run without any termination.

The terminations, basically, it’s based on Spot shortages, right? As long as if you’re using current instances and typical instances, small and medium type of instances without any unique configuration, hardware configuration, then all those instances, typically, they are available in large capacity in AWS. So, those instances will continue to run without any termination, even for weeks and months together.

Spot by NetApp Products

Spot by NetApp, there are various products in Spot by NetApp. What I am talking about is called managed instance, basically stateful instance, how do we provision stateful instance in cloud. So, that’s what we are mainly focusing on here. But there are other products like Cloud Analyzer. Basically, Cloud Analyzer shows the wastages in cloud. It shows what are the workloads you can provision on Spot, or the workloads. It inspects the workloads running on on-demand instances, and it recommends to you which of the workloads are ideal for Spot Instances.

Then there is a product called (Spot) Ocean, where you can leverage Spot Instances when you provision containers using Amazon EKS and ECS kind of platforms. Then, there is the Eco where Spot by NetApp … it looks at your RA usage. And if it is not fully used, then it can even take your RA into RA savings plans into AWS marketplace, and sell or purchase those RAs for you. So, it just optimizes your RAs and savings plan usage. Similarly, we have many other products in Spot by NetApp, but our talk is mainly focused on how do we provision our IaaS VM, stateful VMs in AWS on Spot Instances.

Spot by NetApp Advantage

The thing is, when you take your stateful instances, if you go and provision as it is, if you go and provision your stateful instances on on-demand instances, then your cost actually, especially if you have a large volume of Spot, the VMs, the VM population, then your cloud cost, actually, it’s going to be very, very expensive. In fact, in our case, because of Spot by NetApp and because we could able to migrate our workloads to cloud to Spot instances, our cloud adoption become possible and affordable because of Spot by NetApp.

What are the advantages here? Actually, by using Spot by NetApp, what we get, AWS natively offers the Spot instances, right? But by using Spot by NetApp … what are the advantages there? It’s basically Spot by NetApp, it has some analytical algorithms. It goes and looks into the whole Spot ecosystem. It goes and looks into the Spot availability in each country, each region and each availability zones. And then, it maintains this awareness, right? When you go and provision your application on your Spot instance, you can see the availability, the Spot availability in real time. For instance, the pictures I’m showing here, if you see it on the right-hand side, the scale starts from 0 to 100. One hundred is a strong Spot Instance availability. Zero means there is no Spot Instances there.

Let’s take an example. Let’s take C5 large as an example. So, if you look at it on the left-hand side, the availability zone in US East 1C, C5A large, it’s not highly available. But if you go to US East 1A, it’s highly available there. So, the darker color means it’s highly available, and lesser color means there are less availability there. This way, you get the real-time information. When you even provision your application, the provisioning time itself, Spot can tell you in which availability zone you have more Spot Instances, and where you don’t have any Spot Instance at all.

For instance, on the right-hand side, if you see, we have C3 large here, and then US East 1B, there is no Spot instances available in this EC. There, you get the idea. Okay, let’s go for some other instance type, where there are more Spot instances available… This is just one of the advantage. In the coming slides, we can see Spot helps in a lot many ways to leverage Spot Instances.

Spot by NetApp Advantage (Con’t.)

Here, actually, when you provision your application on the instance, let’s say you are provisioning on your C5 instance, at provisioning time … Spot by NetApp, basically, it suggests the closely matching instances. So, when you provision on your C5 instance, and if there is no C5 in Spot capacity, then you are stuck. Right? So, that’s why it helps you to select closely matching instances.

You can select C4, C5A, C5D … all these instances. You have a few instances you can select, so that there is high availability of Spot instances. When your C5 instance gets terminated, there, it’s highly probable there might be some Spot instance in C4 or C5A or C5D. Spot by NetApp, it goes and picks up another closely matching instance, and then it re-provisions your application there. This is another value add by Spot by NetApp.

Can Spot Instances be leveraged in production?

We talked about when it comes to stateful instances, we know there are these 5 to 10 minutes downtime, the re-provisioning time. Your instance goes away, and then your instance gets re-provisioned. Depending on your instances, it takes 5 to 10 minutes down. So, in that perspective, Spot instances is ideal for stateless containers, because containers, you can drain quickly and re-provision on other available parts quickly. But we can use Spot instances in various other production applications also.

We talked about containers, EKS, ECS, and even on-prem, if you’re using (Red Hat) OpenShift-based container platform. You can leverage Spot instances there, because they’re all stateless. Web servers, also, they are stateless. Since, in a web server farm, you can lose a few web servers, but still you can architect your application in such a way your sessions are redirected to the other web servers. So, there, whether it’s a sub-prod application or production application, Spot instances can be leveraged.

Auto-scaling groups, again, they’re all stateless. Even, actually, Hadoop, like big data applications, there also we can use Spot instances, basically. Because it’s a clustered application, losing a few instances in a cluster, still, your analytical application … it will be running because basically it’s like a clustered application. It ships the workload onto other available cluster nodes. So, even Big Data analytics applications like Hadoop, we can leverage Spot. Even databases, MongoDB and all, if we take, it’s a multi-node replication database, right? If you lose one database, your users are still served from other available databases. There, also, you can use Spot Instance.

Spot by NetApp Adoption at NetApp

Spot by NetApp uses REST API. All your web tools, wherever you are using automation using REST API, you can integrate with Spot by NetApp. It just seamlessly integrates, because it’s a standard REST API based.

Okay. In NetApp, how we are adopting Spot instances, leveraging Spot by NetApp? I will go through each and every item one by one. The left-hand side, what you see, CloudOne IaaS. CloudOne is our self-service automation portal. Here, we provision our IaaS VMs, stateful VMs. What we have done … all our sub-prod applications… we developed automation in such a way if you’re provisioning a sub-prod IaaS VM or stateful VM, it straightaway goes and gets provisioned in Spot Instances. Users, actually, when they provision sub-prod application, it gets directly provisioned on Spot instances. That’s how all our IaaS, our stateful VMs, right now, we started provisioning on Spot instances.


The other platform is CloudOne DevOps. That’s also a self-service portal where developers go and develop their container-based DevOps application. Here, actually, it’s a OpenShift based. Our nodes are basically, earlier, they were running on on-demand instances. Once we started leveraging Spot Instances, we just converted the on-demand instances to Spot Instances. Now, all our container workloads, all the DevOps applications are running on Spot Instances now.

At the bottom, you see the data center right on there. Here, actually, what we are doing, all our application workloads, we are migrating from on-prem data centers to AWS Cloud. So, here, actually, again, all the applications are directly migrated to Spot. This is where we are able to accelerate our cloud adoption, basically because the Spot Instance, it made it affordable for our side. Otherwise, the cloud cost PPC, especially if you have large number of VMs, a thousand VMs, 2,000 VMs kind of population, with VM sizes … large footprint VMs, when you take it to cloud, the cost actually becomes very, very expensive. We were not able to do that, before we started adopting Spot instances, leveraging Spot by NetApp. Now, it becomes possible because, straight away, we get 50% cost reduction, on-demand instances cost. Let’s say if an on-demand instance costs $100… The same instance, if you provision in Spot, it will cost $50 or even less, $40, in that range. So, your cost goes by 50% straight away by just simply provisioning your application on Spot Instances.

Auto-scaling groups, we talked about whatever AWS instances we had, EC2 instances we were running on on-demand. All of them, we converted to spot. We’re also actually, in fact, migrating from OpenShift to Rancher-based container management platform. Spot nicely integrates with Rancher. In AWS, let’s say if we have Amazon EKS and Amazon ECS-based container applications, leveraging Ocean as Spot by NetApp can provision all your EKS containers, ECS containers into Spot instances, and it manages the entire scaling by itself.

We also started looking at the analytical and the multi-master, multi-application databases also. Basically, wherever we are provisioning our applications, we started considering Spot. So, we evaluate whether Spot can be a option there. If we can leverage Spot, then we leverage.

Spot by NetApp Adoption in IaaS (Stateful VMs)

This is the data center application migration from on-prem data center to Spot. Here, actually, we have the data center in RTP, not thoroughly now. There, we are running a hundred plus applications, and thousands of VMs. So, we already started migrating these workloads to AWS, leveraging Spot Instances, using Spot by NetApp. This is a multiyear migration program. We already migrated 28 plus applications and 100 plus VMs there. Again, (in) this journey, we (were) able to migrate our on-prem workload to cloud and reduce our data center footprint. It’s mainly because of Spot Instances, and Spot by NetApp basically helps us there to automate the whole process.

Spot by NetApp Adoption in Stateless DevOps Container Environment

In DevOps area, as I mentioned earlier, our DevOps is Red Hat OpenShift based. Earlier, we have provisioned all our container applications on on-demand EC2 instances. So, what we have done, simply, we converted the EC2 instances to Spot Instances. So, you can just import the instance and then import an EC2 on-demand instance to your Spot instance. It’s that simple. So, we were running about 56 applications there. And once we migrated these worker nodes from on-demand instances to Spot instances, all these were 56 applications we just drained from the on-demand instances, and we provisioned on Spot instances. And our migration from on-demand instances to Spot instances is completed. Basically, all the 56 applications running on Spot instances, and straightaway, we achieved a 50% cost reduction there.

 Production Applications Migrated to Spot (Sample)

Application-wise, I’m just showing some samples here. It’s all business application. So, production applications, it’s running in Spot where some of them are tier two, three, four applications. Where, despite 10 minutes downtime, we can afford. So, if you see some HR portal and NetApp learning applications, those real production applications are running in Spot now.

Production Applications Migrated to Spot (Sample 2)

Sub-prod applications, also, some samples here, some POC machines, Wiki, web pages, again, HR applications, and Unix management servers. In sub-prod ways, there is no constraint at all. Right? All your sub-prod workloads, you can provision on Spot Instances.

Spot Savings in IaaS VMs

Some savings directly taken from the Spot Console, if you see some of the instances. We achieved even 76% savings here. So, your VM, when it costs $100 on on-demand instance, it’s going to cost only $24 when you provision on Spot Instance. So, this is the huge difference, actually. Basically, Spot by NetApp makes the cloud journey our cloud adoption, cloud acceleration, possible.

Stateful IaaS VMs (Logs that show termination and re-provisioning)

So, some logs that shows how a Spot instance gets terminated and relaunched. Here, a termination happens here, and then the relaunching happens here. If you look at the time, it’s just seven, eight minutes the pre-provision gets completed.

Stateful IaaS VMs (Logs that show falling back to OnDemand and then back to spot)

Here, actually, this is one unique feature. If you see here, let’s say, your Spot Instance gets terminated. And then, what do you see here, there are no available matching Spot Instances here. In this case, then Spot by NetApp, what it does, it re-provisions your application on on-demand instance.

That’s what you see here, failing back to on-demand instance, because there is no matching Spot instance. It failed back to on-demand instance. And then, every 15 minutes, it checks for available Spot instances, matching Spot Instances. Once it finds the Spot instance, then it re-provisions, again, from on-demand instance to Spot Instances. It all happens behind the scenes, automatically. So, you don’t need to worry about it.

Basically, what it makes, what Spot by NetApp makes sure, anytime you don’t go out of, let’s say, if there are Spot shortages, your application will not be down more than that 5 to 10 minutes re-provisioning time, right? Either it finds the Spot instance and re-provisions your application on that Spot instance, or it just fail back to an on-demand instance. And in the background, it keeps checking for Spot instances. Once the Spot instance is available, it re-provisions your application from on-demand instances to Spot instances. Basically, at any situation, your application downtime will not go beyond that 5 to 10 minutes.

Spot Migration Demo

Let’s look at how do we provision our application on Spot instances.

Leveraging Spot for Stateful Instances

Here, there are two types. Again, we are talking about only stateful instances here, right? So, there are version like Spot products, where it deals with stateless applications, container-based applications, and etc. Here, actually, our focus is stateful instances. So, I’m going to show how to … You may be running your application on demand instance now. So, how do you import your instance, your application onto Spot instances? That’s a few screenshots I’m going to show you.

The other one is, how do you provision your application fresh using Spot Instances? That is another demo.

Importing an existing stateful instance (Managed instance)

So, first, we look at how to import an existing stateful instance in AWS, which is running on on-demand instances to Spot instances. If you look at here, this is the Spot console. You see here a section called managed instance. We call it the stateful instance as managed instance.

Here, you just go and click a new instance here.

Then, in parallel, you go and look at in AWS your application, which is running on your on-demand instance, right? This is on-demand instance in AWS. You just pick up the instance ID for this on-demand instance. So, you copy the instance ID…

Come back to Spot console, just copy the instance ID here, and select the region, then just click Select.

That’s it.

Since an existing instance, then Spot by NetApp, it goes and finds the instance type, which VPC it’s provisioned, what are the security given groups, AMI image. And all this information, it goes and finds. Since all the instance is already existing and it can find all the instance in all these details from the existing instance, Spot by NetApp just finds out all this information, and then it starts converting that instance as Spot instance.

Basically, it takes a snapshot of that instance and re-provisions that instance, choosing a Spot instance.

It’s so simple, right? Only a couple of mouse clicks, you can just select where existing on-demand instance, and then click on the Import, copy that instance ID, let’s say.

So, in two, three mouse clicks, you will be able to convert all your on-demand instances to Spot instances. Finally, it goes and then it’s the on-demand instance in AWS console.

This is existing Spot Instances. If you want to migrate, it’s so easy to migrate your existing on-demand instances to Spot instances.

Q: What are the commercials and how do you bill partners and the customer?

Tirou: Actually, Spot by NetApp, we charge your portion of your savings, actually. Let’s say you save, I don’t know the exact percentage, it could be 20%, or even less. So, the way we charge is basically we take a portion of your savings. Let’s say $100, the charges there would be $20 or less. I don’t know the exact percentage, how much we are charging, but basically it’s all from your savings. So, the thing is, it’s a safe kind of approach for our customers, right? Only when you save, you pay to NetApp.

Creating a New Spot Instance from Spot Console (Demo)

Moving on. The other type of provisioning… provisioning new application on fresh, earlier, we have seen how to import an application or instance, on-demand instance to Spot instance. But now, we’ll see how to provision your new instance on Spot instance fresh.

Here, actually, all that information, Spot by NetApp, retrieved from your existing instance, you need to give all that data. Since it’s a new instance, you’ll have to provide all that information. That’s the major difference there.

You go to a Spot console and then create a new managed instance.

Then, it asks for the name and then it also asks for the region, which region you want to provision.

…And then, which availability zones you want to provision. Let’s say you select multiple availability zones here. Then, the advantage here is, in US East 1A, your session, your spot, let’s say you have a Spot session and the Spot instance got terminated. Then, if the resources are available in a different AZ, then Spot by NetApp goes and provision your instance on different availability zone.

Here, actually, you just need to be careful. When you go to different availability zones, then your IP address changes, right? There, as long as you are not dependent on your IP address, you don’t need to preserve your IP address. So, here, you can choose multiple AZs. And that’s one way you can highly make your application highly available, because usually a Spot in one AZ, if there is a Spot shortage, then the other AZs most probably might have the same instance type.

Then, you selected your AZ, some region.

And then, you go and select where you want to provision your instance, which VPC, which key pair, the security groups, and which subnet, all these. Earlier, when we import, we don’t need to provide all this information, because it’s all available. Spot by NetApp can find it from the existing instance. But now, since your new fresh instance, you have to go and provision. You have to go and provide all this information there.

Once you have given all that information…

Then you select the OS flavor, the AMI image.

Then, you are selecting the AMI image there.

Actually, when you select the instance type, then this is where you see Spot by NetApp shows the real-time Spot availability. So, it shows whichever region you selected there, and it shows the available AZs, and then in each AZ, what are the Spot availability for the selected instance, the instance type you are choosing. Let’s say here for M1 small, US East 1B, the availability is much less, but the same thing here for T2 micro. You can see it’s highly available.

Depending on that, even if you can fine tune, let’s say, you can choose an instance where the Spot is highly available.

You select your instance and all other parameters here.

Then, at this point in time, Spot by NetApp, it also recommends the closely matching instances. So, it shows some instances. If you want, you can go and add and customize this. So, you can go and add additional instance types here, so that you want to make sure there is always Spot availability. Or you don’t want to give too many options. You can also deselect some of the recommended instance types here, and you can customize this.

There are some startup and shutdown scripts, further customizations here.

Okay. See, I told you, Spot basically takes snapshot of your root volume. So, it re-provisions the root volume. That’s how you get all your applications, everything intact. Spot can also snapshot or backup your EBS volume. So, if you are storing your application data on EBS volumes, Spot by NetApp can backup your EBS volumes. And then, when your instances get terminated, your EBS volumes, if Spot by NetApp re-provisions your application on the same AZ, it just reconnects your EBS volume to the newly provisioned Spot instance. 

Let’s say if your instance gets provisioned on your different AZ, then your EBS volumes on different AZ and your newly provisioned instance on your different AZ. So, what Spot by NetApp … in this scenario, you cannot reattach your EBS volume. Right? Your EBS volume also in that particular AZ, where your AZ information resides. So, in this case, what Spot by NetApp does from the snapshot, it recreates the EBS volume also. So, your instance also get provisioned fresh, and then your EBS volumes also gets provisioned or restored from the snapshot, if your instance is provisioned on a different AZ. It completely takes care, whether it is single AZ restore or if the restore happens on a different AZ. Spot by NetApp can help to restore your application volumes also.

Here, you want to preserve IP. So, you can say you want to preserve IP. Or if your application is not tied to the IP, and if you want to leverage multiple AZs, then you can also use that feature.

So, a feature I wanted to quickly show here, I want to point out here, it’s scheduled action. What happens here, say, you are provisioning your application on Spot instances. Spot instances also helps to schedule your application shutdown and startup here.

You might have heard about the smart parking, right? Let’s say your application, if you want to shut down your application over the weekends, or even nightly, if you want to shut down your application. So, Spot by NetApp can help to schedule that shutdown and startup events also. That is another advantage Spot by NetApp brings in. In fact, what we have done, all our workloads we have migrated to AWS from on-prem. We have first put it on Spot instances. And we are also shutting down those sub-prod instance. Production instances may be production applications. You might want to keep it on 24 hours, right? But sub-prod instances, in our case, we can shut down over weekends. What we have done, we schedule weekend smart parking.

So, in weekend smart parking, what we are doing, Friday evening, 7:00 p.m. or 8:00 p.m., in that range, we shut down our sub-prod instances, and then bring it up on Monday morning, early morning, five o’clock. In the timeframe, we bring it up. … More than 60 hours, we are putting these instances, we keep them in shutdown mode. Here, actually, in a week, when you go for this kind of weekend smart parking, it’s two days out of seven days, right? Straightaway, you can say 25% of your cloud cost, by just putting these instances, just taking these instances into weekend smart parking.

Any questions here?

If you look at the savings, by provisioning your application on Spot instances, you get 50% savings. Then, on top of that, if you do this kind of weekend smart parking, then you get additional 25% savings. Or even if you go for nightly smart parking, then it’s even more savings. Actually, at the end, from your original cost, it’s about 10 to 20% only you will be spending. Almost 80% you can save by just leveraging Spot by NetApp and smart parking.

Once you specify all this information, then just click “Create”. It goes and provisions the new instance on Spot instance.

That brings the end of our presentation. Basically, we are seeing two demos. One is the importing the existing instances. And then, the second demo is how do you provision newer application fresh on Spot instances.

Q: Do you have any metrics on how often Spot instance gets terminated? Is it normally you have it highly available?

Tirou: Actually, what we noticed, as long as if we use current instances … let’s take a CFI family, if you’re using CFI instances, and then if your configuration is regular, normal or typical configuration, such as 4GB, 8GB, 16GB, the common type of CPU and memory configuration… as long as we are using the current version of instances, then the Spot termination, we noticed, it happens very rarely. We have almost 100 plus VMs running in Spot instances now.

From time to time, we see terminations. It varies, essentially, depending on the instance, family instance type. Some application, we see it even run some instances. We see it even runs for one or two months without any interference. But, in some other instance type, we see once in a week, or once in two weeks we see some terminations happens there. But the thing is, you will … never notice, actually, it just gets terminated within that 5 to 10 minutes, your instance gets re-provisioned.

For sub-prod workloads, we don’t get any call from our application developers or users, because the restart or re-provisioning is so quick. We are not getting any calls from users. We hardly notice. So, that’s the advantage there. We can expect there will be terminations, but the impact is very minimal there because Spot by NetApp handles the re-provisioning, and it does it so flawlessly and quickly.

Q: What is the percentage of apps eligible to leverage such Spot by NetApp in a typical IT environment?

Tirou: It’s depends on what type of applications we are trying to put it on Spot. For example, all that stateless applications, containers-based application, you can put it in Spot instances. When it comes to stateful instances, there is no restriction at all. As long as you are able to handle or you are able to tolerate that 5 to 10 minutes downtime, there is no restriction in any application type. So, you can put any application.

Leyddy: Thank you all for attending today. And we’ll be sending the recording and the slide deck shared as well. Also, I will be sending the other recording so that you can have both recordings for the Spot session. And please don’t hesitate to contact me if you have any further questions. Goodbye, everyone. Thank you.