The Six Five On the Road at AWS re:Invent 2022. Patrick Moorhead and Daniel Newman sit down with Shubha Rao, Sr. Manager, Elastic Container Services, AWS.
Their discussion covers:
- Containers (ECS) as a critical part of modernization
- How ECS stacks up in the market
- Real-life examples of ECS deployed
- What can we expect in the future from AWS serverless containers
Be sure to subscribe to The Six Five Webcast so you never miss an episode.
You can watch the full video here:
You can listen to the conversation here:
Disclaimer: The Six Five On the Road is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we do not ask that you treat us as such.
Patrick Moorhead: Hi, this is Pat Moorhead and we are live at AWS Reinvent 2022 in Las Vegas. It’s been an incredible show and we are seeing and talking about some incredible technology. Daniel, how you doing?
Daniel Newman: Pat? Six Five on the road at AWS Reinvent. This is where it’s at. We’ve got multiple days here, a number of great conversations, and we’re going to be taking measurement, taking stock, talking to a lot of executives about what’s going on in the cloud and the enterprise and how technology and IT is proliferating to enable digital transformation.
Patrick Moorhead: Yeah, it’s crazy. The cloud is 15 years old, and 15-year-olds, they’re not the most mature on everything, but they’re in that growth stage where so many people are doing it, and that’s what I’m super excited about. But let’s introduce our guest. Shubha, how you doing?
Shubha Rao: Hi, I’m Shubha. I’m doing good. Thank you for having me here.
Patrick Moorhead: Yeah, it’s great. Maybe, listen, I know you’re famous inside of AWS and with certain audiences, but can you talk about what you do for AWS?
Shubha Rao: Absolutely. So I run the product NDA team in the container services [inaudible 00:01:11] in AWS. So I look after what we call elastic containers, primarily ECS and a few other supporting services.
Patrick Moorhead: I love it. Next generation of application are all pretty much going to be on containers, so a lot of legacy stuff out there. Still some monolithic applications, particularly in finance and retail and, heck, a lot of VMs left as well.
Shubha Rao: Yeah, totally. So, modernization of existing apps and migration to the cloud are our biggest use cases. So, we have a lot of companies across industries trying to use AWS to actually do this, have those journeys, and ECS is one of the primary services that they use.
Daniel Newman: So let’s start there, Shubha. I mean we hear about containers. It’s become really in vogue. Cloud architects and CIOs, just throw in a container and they make it sound so-
Patrick Moorhead: Solved. Check. Right? Everything’s solved.
Daniel Newman: But, it’s kind of a well understood part of the strategy to modernize IT. Talk a little bit about that and how containers became such a big part of it and how AWS is sort of thinking about it.
Shubha Rao: Absolutely. So, containers started as a way to have this consistency between what a developer builds and what you actually have deployed on the production, but from there in fact we have added a lot of features to make the whole application deployment journey easier for customers.
So, today we support a lot of automated scaling and cost management, and we integrate with so many different AWS services to just get a end to end experience in both deploying applications as well as having them grow and scale and be optimized over time. And that’s where we are seeing more customers see the value of container orchestrator systems, and it’s become the application modernization journey of choice today.
Patrick Moorhead: All containers are not built the same, kind of like wine. I mean we say, “Hey, wine, well you want red wine, you want white wine?” I mean, we don’t just stop there. There’s 270 different variations of it just down at the local store. So, can you talk about what separates just a classic Kubernetes container and an ECS?
Shubha Rao: So a container is still the same. Both Kubernetes as well as ECS support deploying the standard container in our packaged applications. The approach we take into what APIs and what interfaces we provide and what integrations we provide out of the box versus what is available for customers to extend is a little bit different.
So ECS is a flavor of container orchestrator where Amazon manages the control plane. Amazon manages most of the extensions and the add-ons or controllers that is required for running applications. So that means you get more out of the box, and in our company’s platform, teams can rely on us to keep it available and running and scaling really well.
And, we manage a lot of the patching, and as we add new features, customers don’t have to do anything additional to get those, and we kind of just give it out of the box. And, that’s the key difference in implementation. The API that a customer gets is different and how managed it is and what part of it is managed by AWS versus by the customer is different. And, ECS approach is to give customers a lot more out of the box than what you get from running a Kubernetes cluster.
Patrick Moorhead: So, with simplicity, does it come cost reduction, less resources, you can do more with the same resources? What’s the end benefit of choosing ECS over another container that quite frankly you could run on AWS?
Shubha Rao: The time to production. How fast can you bring an application and run it on ECS? How easily can you scale and extend it and how much effort do you have to put into keeping that platform available and secure, and cost optimizations and scaling just out of the box. So, these are the three benefits our customers see.
This time to production is one of the biggest things that a lot of the new customers that are new to the cloud really optimize for because they have drivers that include moving off their data centers, or just being able to scale faster based on incoming needs, and that kind of makes them choose a platform where a lot of that is handled for them as opposed to them actually having to do the testing and all the implementation and testing related to that.
Patrick Moorhead: Right. Sounds like the way that I would like my containers. Simple, please.
Daniel Newman: Or your wine.
Patrick Moorhead: And watches, too
Daniel Newman: You kind of talk about simplicity as a case, and as analysts we love to look at the crawl and the charts, and we like data, so we love to do research and reports. Any sense of ease of use? Any sense from those interactions you have with customers? Is it shortening things by percentages, days, hours, weeks? Share a little bit about some of those findings.
Shubha Rao: So, in this reinvent we have markers team from Goldman Sachs who sharing their journey about how they were able to launch new financial applications faster. So before, they were running a legacy platform where it was taking a lot of time and effort for them. When they actually chose ECS Fargate as the platform of choice, they were able to launch multiple new credit card and other offerings within months as opposed to the years of planning that it used to take them in the past.
We have customers like Medlife, who is a healthcare firm who actually was initially deploying new applications in months, and it used to take them two months to get new feature out and running and available to their customers, and now they can do multiple launches every day. Their developer teams are a lot more agile, and they’re able to serve up to a 10X increase in traffic just because they have a lot of that handled for them by ECS.
We also have a lot more financial services who have gone all in on ECS just because of the security posture that ECS provides out of the box. Companies like Capital One and Vanguard are almost exclusively ECS Fargate first. So they have thousands of applications that they’ve modernized as well as they’ve migrated to the cloud and they now have multiple deployments every day. They have tens of thousands of microservices, all of them running at a fraction of the cost and with a big improvement in their developer productivity just because they’re able to automate a whole bunch of things.
Daniel Newman: So, you leaned into financial services. It sounds like that’s been a big win, and that’s kind of interesting, too, because a lot of times people sometimes think financial services don’t move as fast to the cloud because of compliance and regulatory governance. Of course, they’re moving, but certain workloads, certain privacy, certain data, that sounds like a very strong use case in industry. Any other industries besides financial services that you’re seeing kind of with similar momentum around ECS?
Shubha Rao: Yeah, we are seeing it from a variety of industries. So, from healthcare, one example is a company called Health One, which was a recent win. They were doing Covid testing and, as you can expect, the increase in demand for the test kits and how quickly the results be shared with the customers was all very critical for them, and they were able to actually scale up to 35 million tests just all handled on the ECS platform, and they were able to handle it without really having to make a major investment into how they build and run applications. That’s one example for there.
We also have a lot of consumer applications. For example Duolingo, all of us have tried to use that. They had, again, in the last two, three years, their traffic has increased multifold, and they’re actually able to achieve a cost reduction in addition to serving their multifold increase in traffic because they’re able to leverage things like capacity for orders and using spot capacity and optimizing to actually only use the CP on memory when they actually have requests coming in versus all the time, and they’re able to run tens of microservices that are all interconnected with each other with this type of architecture.
Patrick Moorhead: How do your customers determine whether, let’s say it’s new application, and maybe we can talk about moving from VMs to ECS as a follow up, but how do you guide them? Do I use ECS, do I use EKS, do I use Lambda and go serverless? I mean, what’s the breakdown? Is there a decision tree of what to pick?
Shubha Rao: Yes. So, we should be speaking more about our decision tree, but our guidance to customers is to choose the highest level of abstraction that they can be happy with for their application. So, if your application can actually, if it is just code that you bring and we handle everything for you, including integration, scaling, everything else, then Lambda and App Runner are your first choice. That’s general. That gives you the most serverless experience, most of the integrations provided out of the box.
If you need some extensions, some extensibility including having access to your data, to your compute just because you have a compliance or a mandate like security mandate to actually run your own agents, that’s when ECS is a good fit for you, where it kind of provides a mixture of extensibility and simplicity out of the box.
So, we have partner integrations, we have tens of partners who are working with us to actually give flexibility for customers and those extension points are some things that our customers used to build their own developer portals and their own dashboards for assessing different types of data, but you still have the control plane managed by us, scaling managed by us and all the integrations managed by us.
And, it’s a little bit of a hands on approach. Ongoing operations is much lesser with ECS, and that’s our guidance. There are many of our customers who have the need to keep their platform really customized to their applications, and that’s when we actually recommend going the EKS route, where EKS gives you the standard Kubernetes APIs and all the extension that Kubernetes supports is something that the customers have flexibility to use and leverage.
Patrick Moorhead: Great answer. No, no, this is good. So with so many applications on VMs, do you guide customers on when to move, don’t move, and I know you can mix containers and VMs just to be on the, I don’t know, not on the safe side, but to change less. How does that work, and do you see a lot of them moving over?
Shubha Rao: We see a lot of customers moving over. Again, customers also have multiple compute platforms. Most of our customers are running something in EC two, something in containers, something in functions, and they choose the right platform for their needs based on their applications requirements, and that’s the approach that we recommend as well just to get the simplicity and performance and cost optimized based on what the application requires.
Daniel Newman: It’s interesting, because we kind of keep bouncing between simplicity and complexity. When I listen to you talk, I start to actually feel overwhelmed by the complexity. Well, because like you said, these environments tend to be more heterogeneous than… So it’s never like we’re just going to do everything in ECS. It’s like we’ve got a combination of where our compute exists, we’ve got different data and storage and locations, we’ve got Edge. And all these things are trying.
So, I guess kind of a big question, because obviously ECS sounds like it’s the most comprehensive in terms of the offering.
Patrick Moorhead: It’s easy button, too.
Daniel Newman: Yeah, it’s the easy button in a lot of ways. But, how much is managing complexity? And let me just, I’m hearing more and more that containers was supposed to be the catchall solution for a lot of companies, and this is how enterprise IT was going to get a handle on their IT environments. And now we’re starting to hear that a lot of these people that went all in on containers are going back kind of like, “Whoa, we took on a lot,” and the SI are becoming more important, the consultants and they’re kind of almost re-looking at their strategies.
Shubha Rao: So, I want to distinguish between containers complexity versus container platform and orchestration system complexity. So, containers is still what we see a lot of customers adopt, and that may be something that customers directly use in their developer environment and then bring it to cloud environment, or they can bring code and we will containerize and run it just because the right or the best practice as of today is to actually run containers so that they get the consistency on other things.
So, our portfolio of services supports the whole customers bring your business logic, and we do the whole containerization and applying a whole bunch of best practices all from management to secrets management to logging and observability and compliance and any other thing that customer, like firewalls and other things that customers needs.
And that’s a service that we have called App Runner. And that whole concept of end to end migration management of running your application is the implementation detail that is containers, and there are other approaches including ECS and EKS, which actually give customers, it’s like peeling the onions, a little more flexibility, and a little more flexibility, and not every customer requires all of that flexibility, and it’s their choice to decide how much of it is something they’re taking on.
Because with all of that flexibility comes a lot of the operations and management, ongoing management. So, that’s our trade off and the choice that our customers need to make for themselves. And many of our customers are trying out multiple things and actually moving between each other, which is not uncommon. They start with a lot of what they hear from people like you, and they start somewhere, but then over time-
Patrick Moorhead: Well, we generally, we simplify things to the point of reality, and that’s why we love talking to real enterprises to get the real deal on what’s going on. And, it is tough, but the great thing is different strokes for different folks. Wherever you are even on your journey, the level of portability you want versus maybe the easy button inside of a certain environment, I think that’s good for… In the end I think it’s good for customers, and I think it’s good for the industry as a whole. So, yeah.
Shubha Rao: One more point, you spoke about Edge and anywhere, so that’s actually another trend that we are noticing. A lot of customers want Amazon to help in migrating their Edge applications or managing their Edge applications as a part of our service. So, ECS provides a capability called ECS anywhere, which is you use the same on cloud management and APIs and the application management experience. You can run your tasks and services and have it integrated with monitoring tools, but you can also deploy that same app not just in region, in an AWS region, but on a customer data center or on a customer Edge device or a branch office where they have servers.
Patrick Moorhead: But, you can actually run it on an Edge device.
Shubha Rao: You can. I am not really running-
Patrick Moorhead: You that in banking, where they have to keep compliance, and they have to have essentially you’re tunneling in to that. So, okay, I’m figuring this out here, and I promised I wouldn’t say EKS.
Daniel Newman: So that’s why she said layers of abstraction. I mean they’re-
Patrick Moorhead: Yeah.
Shubha Rao: EKS does provide EKS Anywhere where you’re running the entire stack in a isolated or non-isolated environment. ECS provides a different approach to managing anywhere, managing applications anywhere, where you’re not deploying and managing the entire stack, you’re basically connecting your Edge location to the cloud and using the cloud control to actually get the latest software deployed out into the remote offices. And, that’s just a simplicity of managing.
And also, that gives a easy path to burst your applications to the cloud when you need additional capacity or to actually migrate gradually when being able to test both environments and scaling on to the cloud if needed.
Patrick Moorhead: So, we have covered almost the gamut. I mean I threw the monolithic application thing out there, we talked about VMs, we talk ECS, EKS, running things everywhere, the simple button. Maybe we talk about the future and either your vision or the way that you think the industry at a whole is going to pan out in the future.
Shubha Rao: Cost optimization is top of mind for a lot of our customers, and we are bringing in new features to help customers understand cost, not just is my cost container using level, but what’s my service, what’s my environment? Which team is deploying how many applications, and what’s the cost of that? So, that’s one area that we are investing into.
Patrick Moorhead: What’s an example of a… Again, I’m not asking you to pre-announce something, but what’s something that could make a more abstraction.
Shubha Rao: So, ECS provides a capability that’s like called service where you run end copies of an application just based on being able to serve the number of requests that you are getting for your application. So, traditionally you had to, if you have service A connecting to service B, somebody had to deploy a load balancing service and have them connect with each other. And it could be for an external facing service.
A load balancer is required in a lot of ways, but if you have hundreds of internal services deploying all of that, it’s kind of shifting the mental model of development teams from thinking about just their application to thinking about the infrastructure and target group and with what it’s connected with. One of the new launches invent from the ECS team is called Service Connect, which is bringing the connectivity experience integrated into the service construct.
So, you can say service A with this friendly name, the front end can connect to service B, my middleware, my backend without really having to think through or deploy any other infrastructure. So, underneath the scene we install a set of things, we configure a set of things, but customers can actually get that simpler and a higher abstraction that helps them think develop application first as opposed to having to think infrastructure first.
And, this is another thing that we have in our vision and roadmap where we want to make it simpler for customers to think about amendments and to think about they just could stay in the IDE and we can do everything from taking the code there to building the automation required for deploying and containerizing and in integrating other frequently used services and applying into an environment and scaling it on their behalf.
Daniel Newman: So we have just a few minutes left and you said something that tickled my fancy. I love talking about cloud economics, and you said cost optimization’s a big part of it.
Patrick Moorhead: I knew you’d drill down.
Daniel Newman: Yeah, I can’t help it. Okay, so let’s start with the optimization thing. First is a simple question. Are you seeing a lot more interest in the last quarter, three months, six months as we’ve seen the economy? Is that kind of what you think’s driving that conversation?
Shubha Rao: Cost is always top of mind for customers, and it’s become more top-
Daniel Newman: Okay. Yeah, I was going to say but are we seeing that spike? Because we’re hearing more stories about the clouds getting a closer look. So, can you talk a little bit about your managing portfolio, how does that really yield cost optimization in the customer conversations you’re having?
Shubha Rao: Yeah, so ECS provides integration with Fargate. Fargate is a way to just pay for the capacity that your tasks are using as opposed to provisioning a set of VMs and paying. And, thinking about the optimization of that cluster and thinking about how to keep it optimal so that you’re able to scale but also not overpaying for it.
So, Fargate gives you a pay per use CPU memory based model to running applications, which is the first step that we took five years ago. And over time we’ve added Spark, we have added capacity, a construct called Capacity Provider, which helps you rebalance between the EC two that you have probably reserved, and paying a reduced discounted price for. And Fargate and rebalancing your tasks across multiple types of capacity that plug in into ECS.
And over time we’ll actually have more automated and intelligent capacity optimization, which is something on a roadmap. And also rebalancing your tasks so that you can… You don’t have to pre-provision, and we do the whole unscaling or scaling it down when there is not really much traffic that applications are seeing.
Patrick Moorhead: Yeah, it’s fascinating. Ten years ago, AWS customers would get an email and say, “Hey,” we call it downshifting. And, gosh, just another example of things actually coming true that we say, which is then getting to automation, which customers, in the end, they want. And I think that engenders trust in you in that, heck, the revenue might go down for somebody, but I think it engenders so much trust, “Hey I want to give more, I want to do more with AWS.”
And I think the rest of the industry really needs to think through that, because it is a long term customer satisfier that just keeps people coming back and doing more. And, I think, I look across it the last 30 years, and that’s brand new mindset. And, I remember the good old days, too, with when S3 and it was, “Hey, every month we’re going to have a price decrease on this.” And it kind of got this notion of, “Okay, I can bank on this. This isn’t going to be a we’re going to come in, they’re going to lower their costs, and my costs aren’t going to lower.” So bravo on that.
I mean, some things you have to keep in mind, obviously, with Fargate is the timing of the switching between downshifting and automatic up shifting that people talk about. It’s the same thing for Lambda as well. I mean you just have to have the right type of application that works in that environment. But bravo on the auto shift there.
Shubha Rao: Thank you.
Daniel Newman: Well Shubha, I want to thank you so much for joining us here on the Six Five, on the road at AWS reinvent 2022. It’s great having you on. Hopefully, you’ll be coming back onto the show. We’ll have you back again next year. I want hear the advancements of ECS, a very important technology. Enterprises are definitely leaning into this container strategy right now, and it sounds like you’re right at the forefront of it.
Shubha Rao: Yeah, thank you for having me. Last data point I want to share, more than 65% of the containers customers choose ECS as their first service and deploy on ECS, and we are able to support the largest workload up to more than 5 billion hours, and we are able to support super fast scaling. So, try it out and let us know how you like it and what you need from us.
Patrick Moorhead: That’s great.
Daniel Newman: It started with that data point.
Patrick Moorhead: I was thinking the exact same thing. No, no this is great. This is a super strong ending. No, thanks so much Shubha for coming on.
So, we hope you enjoyed this discussion on ECS, and we’re going to have a lot of videos this week talking to some pretty incredible people and some pretty incredible technology here at AWS Reinvent 2022. If you like what you heard, hit that subscribe button and we just thank you, and have a good morning, afternoon, night, wherever you are on the planet. Take care.