Podcast: The surging role of containers in the 'hybrid estate'
[Editor's note: This podcast was originally published on Sept. 30, 2019.]
Over the past decade or so, computing environments have moved from bare metal to virtualization, enabling greater agility, among other advantages. But virtualization has been “fairly expensive,” says HPE’s Mark Linesch, “because you are virtualizing the entire infrastructure, if you will, and dragging along a unique operating system each time you do that.”
So, what comes next? In today’s hybrid world, where computing is distributed from edge to cloud, containerization is increasingly the answer, Linesch says―particularly for developers. Using containers, developers can “couple not only the application but the application dependencies, the libraries, and all that they need to run an application throughout the DevOps lifecycle,” he explains. In addition to enabling more flexibility and faster time to deployment, container-based hybrid approaches are driving increased automation, built-in intelligence, and robust management of data and workloads.
In this Hewlett Packard Enterprise Voice of the Innovator podcast, Linesch and host Dana Gardner explore the latest in containerization, including how standards, microservices, composable infrastructure, and more play into the equation.
The following are excerpts from the discussion.
Gardner: Let’s look at the state of the industry around containers. What are the top drivers for containers adoption now that the technology has matured?
Linesch: The history of computing, as far back as I can remember, has been about abstraction, abstraction of the infrastructure, and then a separation of concern between the infrastructure and the applications.
Mark Linesch, HPE
It used to be it was all bare metal, and then about a decade ago, we went on the journey to virtualization. And virtualization is great; it’s an abstraction that allows for a certain amount of agility. But it’s fairly expensive because you are virtualizing the entire infrastructure, if you will, and dragging along a unique operating system (OS) each time you do that.
So the industry for the last few years has been saying, “Well, what’s next—what’s after virtualization?” And clearly things like containerization are starting to catch hold.
Why now? Well, because we are living in a hybrid cloud world, and we are moving pretty aggressively toward a more distributed edge-to-cloud world. We are going to be computing, analyzing, and driving intelligence in all of our edges—and all of our clouds.
Things such as performance- and developer-aware capabilities, DevOps, the ability to run an application in a private cloud and then move it to a public cloud, and being able to drive applications to edge environments on a harsh factory floor—these are all aspects of this new distributed computing environment that we are entering into. It’s a hybrid estate, if you will.
Containers have advantages for a lot of different constituents in this hybrid estate world. First and foremost are the developers. If you think about development and developers in general, they have moved from the older, monolithic and waterfall-oriented approaches to much more agile and continuous integration and continuous delivery models.
And containers give developers a predictable environment wherein they can couple not only the application but the application dependencies, the libraries, and all that they need to run an application throughout the DevOps lifecycle. That means from development through test, production, and delivery.
Containers carry and encapsulate all of the app’s requirements to develop, run, test, and scale. With bare metal or virtualization, as the app moved through the DevOps cycle, I had to worry about the OS dependencies and the type of platforms I was running that pipeline on.
Developers’ package deal
A key thing for developers is they can package the application and all the dependencies together into a distinct manifest. It can be version-controlled and easily replicated. And so the developer can debug and diagnose across different environments and save an enormous amount of time. So developers are the first beneficiaries, if you will, of this maturing containerized environment.
But next are the IT operations folks because they now have a good separation of concern. They don’t have to worry about reconfiguring and patching all these kinds of things when they get a hand-off from developers into a production environment. That capability is fundamentally encapsulated for them, and so they have an easier time operating.
And increasingly in this more hybrid distributed edge-to-cloud world, I can run those containers virtually anywhere. I can run them at the edge, in a public cloud, in a private cloud, and I can move those applications quickly without all of these prior dependencies that virtualization or bare metal required. It contains an entire runtime environment and application, plus all the dependencies, all the libraries, and the like.
The third area that’s interesting for containers is around isolation. Containers virtualize the CPU, the memory, storage network resources—and they do that at the OS level. So they use resources much more efficiently for that reason.
Unlike virtualization, which includes your entire OS as well as the application, containers run on a single OS. Each container shares the OS kernel with other containers, so it’s lightweight, uses much fewer resources, and spins up almost instantly—in seconds versus virtual machines (VMs) that spin up in minutes.
When you think about this fast-paced, DevOps world we live in—this increasingly distributed hybrid estate from the many edges and many clouds we compute and analyze data in—that’s why containers are showing quite a bit of popularity. It’s because of the business benefits, the technical benefits, the development benefits, and the operations benefits.
Gardner: It’s been fascinating for me to see the portability and fit-for-purpose containerization benefits, and being able to pass those along a DevOps continuum. But one of the things that we saw with virtualization was that too much of a good thing spun out of control. There was sprawl, lack of insight and management, and eventually waste.
How do we head that off with containers? How do containers become manageable across that entire hybrid estate?
Setting the standard
Linesch: One way is standardizing the container formats, and that’s been coming along fairly nicely. There is an initiative called The Open Container Initiative, part of the Linux Foundation, that develops to the industry standard so that these containers, formats, and runtime software associated with them are standardized across the different platforms. That helps a lot.
Number two is using a standard deployment option. And the one that seems to be gripping the industry is Kubernetes. Kubernetes is an open source capability that provides mechanisms for deploying, maintaining, and scaling containerized applications. Now, the combination of the standard formats from a runtime perspective with the ability to manage that with capabilities like Mesosphere or Kubernetes has provided the tooling and the capabilities to move this forward.
Gardner: And the timing couldn’t be better, because as people are now focused on as-a-service for so much—whether it’s an application, infrastructure, and increasingly, entire data centers—we can focus on the business benefits and not the underlying technology. No one really cares whether it’s running in a virtualized environment, on bare metal, or in a container—as long as you are getting the business benefits.
Linesch: You mentioned that nobody really cares what they are running on, and I would postulate that they shouldn’t care. In other words, developers should develop, operators should operate. The first business benefit is the enormous agility that developers get and that IT operators get in utilizing standard containerized environments.
Not only do they get an operations benefit, faster development, lower cost to operate, and those types of things, but they take less resources. So containers, because of their shared and abstracted environment, really take a lot fewer resources out of a server and storage complex, out of a cluster, so you can run your applications faster, with less resources, and at lower total cost.
This is very important when you think about IT composability in general because the combination of containerized environments with things like composable infrastructure provides the flexibility and agility to meet the needs of customers in a very time sensitive and very agile way.
Gardner: How are IT operators making a tag team of composability and containerization? Are they forming a whole greater than the sum of the parts? How do you see these two spurring innovation?
Linesch: I have managed some of our R&D centers. These are usually 50,000-square-foot data centers where all of our developers and hardware and software writers are off doing great work.
And we did some interesting things a few years ago. We were fully virtualized, a kind of private cloud environment, so we could deliver infrastructure-as-a-service (IaaS) resources to these developers. But as hybrid cloud hit and became more of a mature and known pattern, our developers were saying, “Look, I need to spin this stuff up more quickly. I need to be able to run through my development-test pipeline more effectively.”
And containers as a service was just a super hit for these guys. They are under pressure every day to develop, build, and run these applications with the right security, portability, performance, and stability. The containerized systems—and being able to quickly spin up a container, to do work, package that all, and then move it through their pipelines—became very, very important.
From an infrastructure operations perspective, it provides a perfect marriage between the developers and the operators. The operators can use composition and things like our HPE Synergy platform and our HPE OneView tooling to quickly build container image templates. These then allow those developers to populate that containers-as-a-service infrastructure with the work that they do—and do that very quickly.
Gardner: Another hot topic these days is understanding how a continuum will evolve between the edge deployments and a core cloud, or hybrid cloud environment. How do containers help in that regard? How is there a core-to-cloud and/or core-to-cloud-to-edge benefit when containers are used?
Gaining an edge
Linesch: I mentioned that we are moving to a much more distributed computing environment, where we are going to be injecting intelligence and processing through all of our places, people, and things. And so when you think about that type of an environment, you are saying, “Well, I’m going to develop an application. That application may require more microservices or more modular architecture. It may require that I have some machine learning (ML) or some deep learning analytics as part of that application. And it may then need to be provisioned to 40—or 400—different sites from a geographic perspective.”
When you think about edge-to-cloud, you might have a set of factories in different parts of the U.S. For example, you may have 10 factories all seeking to develop inferencing and analyzed actions on some type of an industrial process. It might be video cameras attached to an assembly line looking for defects and ingesting data and analyzing that data right there, and then taking some type of a remediation action.
And so as we think about this edge-to-cloud dance, one of the things that’s critical there is continuous integration and continuous delivery—of being able to develop these applications and the artificial intelligence (AI) models associated with analyzing the data on an ongoing basis. The AI models, quite frankly, drift and they need to be updated periodically. And so continuous integration and continuous delivery types of methodologies are becoming very important.
Then, how do I package up all of those application bits, analytics bits, and ML bits? How do I provision that to those 10 factories? How do I do that in a very fast and fluid way?
That’s where containers really shine. They will give you bare-metal performance. They are packaged and portable, and that really lends itself to the fast-paced delivery and delivery cycles required for these kinds of intelligent edge and Internet of Things (IoT) operations.
Gardner: We have heard a lot about AIOps and injecting more intelligence into more aspects of IT infrastructure, particularly at the June HPE Discover conference. But we seem to be focusing on the gathering of the data and the analysis of the data, and not so much on the what do you do with that analysis—the execution based on the inferences.
It seems to me that containers provide a strong means when it comes to being able to exploit recommendations from an AI engine and then doing something—whether to deploy, to migrate, to port.
Am I off on some rough tangent? Or is there something about containers—and being able to deftly execute on what the intelligence provides—that might also be of benefit?
Linesch: At the edge, you are talking about many applications where a large amount of data needs to be ingested. It needs to be analyzed, and then take a real-time action from a predictive maintenance, classification, or remediation perspective.
And so containers spin up very quickly. They use very few resources. The whole cycle time of ingesting data, analyzing that data through a container framework, taking some action back to the thing that you are analyzing is made a whole lot easier and a whole lot performant with less resources when you use containers.
Now, virtualization still has a very solid set of constituents, both at the hybrid cloud and at the intelligent edge. But we are seeing the benefits of containers really shine in these more distributed edge-to-cloud environments.
Gardner: Mark, we have chunked this out among the developer to operations and deployment, or DevOps implications. And we have talked about the edge and cloud.
But what about at the larger abstraction of impacting the IT organization? Is there a benefit for containerization where IT is resource-constrained when it comes to labor and skills? Is there a people, skills, and talent side of this that we haven’t yet tapped into?
Customer microservices support
Linesch: There definitely is. One of the things that we do at HPE is try to help customers move into these new models like containers, DevOps, and continuous integration and delivery. We offer a set of services that help customers, whether they are medium-sized customers or large customers, to think differently about development of applications. As a result, they are able to become more agile and microservices-oriented.
Microservice-oriented development really lends itself to this idea of containers and the ability of containers to interact with each other as a full-set application. What you see happening is that you have to have a reason not to use containers now.
That’s pretty exciting, quite frankly. It gives us an opportunity to help customers to engage from an education perspective and from a consulting, integration, and support perspective as they journey through microservices and how to rearchitect their applications.
Our customers are moving to a more continuous integration-continuous development approach. And we can show them how to manage and operate these types of environments with high automation and low operational cost.
Gardner: A lot of the innovation we see along the lines of digital transformation at a business level requires taking services and microservices from different deployment models—oftentimes multicloud, hybrid cloud, SaaS, on-premises, bare metal, databases, and so forth.
Are you seeing innovation percolating in that way? If you have any examples, I would love to hear them.
Linesch: I am seeing that. You see that every day when you look at the Internet. It’s a collaboration of different services based on APIs. You collect a set of services for a variety of different things from around these Internet endpoints, and that’s really as a service. That’s what it’s all about—the ability to orchestrate all of your applications and collections of service endpoints.
Furthermore, beyond containers, there are new as-a-function-based, or serverless, types of computing. These innovators basically say, “Hey, I want to consume a service from someplace, from an HTTP endpoint, and I want to do that very quickly.” They very effectively are using service-oriented methodologies and the model of containers.
We are seeing a lot of innovation in these function-as-a-service (FaaS) capabilities that some of the public clouds are now providing. And we are seeing a lot of innovation in the overall operations at scale of these hybrid cloud environments, given the portability of containers.
At HPE, we believe the cloud isn’t a place—it’s an experience. The utilization of containers provides a great experience for both the development community and the IT operations community. It truly helps better support the business objectives of the company.
Investing in intelligent innovation
Gardner: Mark, for you personally, as you are looking for technology strategy, how do you approach innovation? Is this something that comes organically, that bubbles up? Or is there a solid process or workflow that gets you to innovation? How do you foster innovation in your own particular way that works?
Linesch: At HPE, we have three big levers that we pull on when we think about innovation.
The first is we can do a lot of organic development, and that’s very important. It involves understanding where we think the industry is going and trying to get ahead of that. We can then prove that out with proof of concepts and incubation kinds of opportunities with lead customers.
We also, of course, have a lever around inorganic innovation. For example, you saw recently an acquisition by HPE of Cray to turbocharge the next generation of high-performance computing (HPC) and to drive the next generation of exascale computing.
The third area is our partnerships and investments. We have deep collaboration with companies like Docker, for example. They have been a great partner for a number of years, and we have, quite frankly, helped to mature some of that container management technology.
We are an active member of the standards organizations around the containers. Being able to mature the technology with partners like Docker, to get at the business value of some of these big advancements, is important. So those are just three ways we innovate.
Longer term, with other HPE core innovations, such as composability and memory-driven computing, we believe that containers are going to be even more important. You will be able to hold the containers in memory-driven computing systems, in either dynamic random-access memory (DRAM) or storage-class memory (SCM).
You will be able to spin them up instantly or spin them down instantly. The composition capabilities that we have will increasingly automate a very significant part of bringing up such systems, of bringing up applications, and really scaling and moving those applications to where they need to be.
One of the principles that we are focused on is moving the compute to the data, as opposed to moving the data to the compute. And the reason for that is when you move the compute to the data, it’s a lot easier, simpler, and faster—with less resources.
This next generation of distributed computing, memory-driven computing, and composability is really ripe for what we call containers in microseconds. And we will be able to do that all with the composability tooling we already have.
Gardner: When you get to that point, you’re not just talking about serverless. You’re talking about cloudless. It doesn’t matter where the FaaS is being generated as long as it’s at the right performance level that you require, when you require it. It’s very exciting.
Before we break, I wonder what guidance you have for organizations to become better prepared to exploit containers, particularly in the context of composability and leveraging a hybrid continuum of deployments. What should companies be doing now in order to be getting better prepared to take advantage of containers?
Be prepared, get busy
Linesch: If you are developing applications, then think deeply about agile development principles, and developing applications with a microservice-bent is very, very important.
If you are in IT operations, it’s all about being able to offer bare metal, virtualization, and containers-as-a-service options—depending on the workload and the requirements of the business.
I recommend that companies not stand on the sidelines but to get busy, get to a proof of concept with containers as a service. We have a lot of expertise here at HPE. We have a lot of great partners, such as Docker, and so we are happy to help and engage.
We have quite a bit of onboarding and helpful services along the journey. And so jump in and crawl, walk, and run through it. There are always some sharp corners on advanced technology, but containers are maturing very quickly. We are here to help our customers on that journey.
Role of containers in the 'hybrid estate': Lessons for leaders
- Containers are gaining steam in today's hybrid world because they can run at the edge, in a public cloud, in a private cloud—virtually anywhere.
- Efforts such as the Linux Foundation's Open Container Initiative are developing industry standards so containers, formats, and the runtime software associated with them are standardized across the different platforms.
- Going forward, container-based approaches will incorporate artificial intelligence, microservices, composability, and memory-driven computing.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.