Why devops needs a better approach to cloud networking

• Read original article here

The promise of digital transformation is enabling businesses to magnify competitive advantages, create new revenue streams, and improve customer experiences. To accomplish all of this, devops engineers are asked to build, test, and deploy applications and services with increased velocity using Kubernetes, microservices, and other cloud-native computing tools, backed by resources from cloud service providers depending on preferences and requirements.

But just as your engineers and application stack must turn more agile, so must your network. Shifting to full-stack, autonomous networking for multicloud environments offers a way for enterprises to overcome time-to-value issues and unleash devops engineers to maximize productivity and contributions to business growth. 

While delivering applications and services more quickly offers huge business benefits, it also comes with added challenges and risks. What good is an application if users experience performance problems that make them less productive? Are your applications providing a secure experience or exposing the organization, employees, or customers to risk? Have all the regulatory compliance requirements been met? 

More applications are moving to the cloud, and it is expected that cloud spending will outpace non-cloud IT infrastructure spending for the first time later this year, according to IDC. Furthermore, Prosimo’s recent State of Multi-Cloud Infrastructure Report found that 91% of enterprises plan to use multiple clouds, 62% within two years. With greater scale comes increased complexity, and enterprises are struggling to orchestrate and manage this new dynamic IT environment across on-premises data centers, edge computing, and cloud footprints in a consistent way. 

Many enterprises have tackled connectivity requirements using traditional legacy networking tools, but that approach has fallen short. In fact, 53% of the enterprises tackling multicloud with traditional networking approaches face issues with operational complexity, security, and performance, according to the same State of Multi-Cloud Infrastructure Report. This is due to several key issues.

First is that, with traditional networking approaches, troubleshooting and resolving problems is a time-consuming, manual process of looking at dozens of different networking and security tools and dashboards. Second, dealing directly with the native networking tools from Amazon Web Services, Microsoft Azure, Google Cloud Platform, and other cloud providers means navigating different terminology, different user interfaces, and different features, adding more complexity. And third, traditional networking tools have blind spots, as they don’t look at the application layers to provide the best user experience.

Enterprises need to solve connectivity and cloud migration problems, but that’s only the start. They also need to think beyond connectivity so that they can free up devops resources to create a frictionless process for delivering applications.

An attractive and increasingly popular way to create dynamic networking platforms is to build on cloud-native constructs related to application and connectivity requirements. Cloud-native constructs give enterprises a way to build scalable platforms that can simplify the orchestration and management of cloud (and on-premises) resources, including integration of native functionality and new features from cloud providers. Cloud-native constructs also make it seamless to give enterprises end-to-end visibility of the entire cloud footprint in real time. 

While legacy networking and security tools provide point-in-time information for a specific scenario, they often don’t provide the complete picture. Cloud-native constructs offer a way for enterprises to dynamically build a consistent, full-stack networking architecture that reduces the complexity of connecting applications, services, and networks across clouds. Think of full-stack as having two core components. One layer focuses on connectivity and scales dynamically, like hyperscalers, taking advantage of the cloud backbone. The second layer, which sits on top, understands the application and how to best interconnect to ensure that security and performance knobs are available.

Delivered as a single, integrated architecture, a full-stack networking platform addresses operations and interdependencies across the full lifecycle—from the network to the application layer. It accounts for networking, performance, security, compliance, and even cloud spend.

With complete end-to-end visibility from a full-stack networking platform, enterprises can start to build policies for common issues and requirements. Of course, enterprises are constantly evolving due to economic pressures, shifting customer needs, and new market opportunities. Machine learning offers a way to more easily adapt to new business requirements.

Machine learning can recognize patterns to identify issues and provide recommendations to resolve them. Machine learning helps enterprises resolve many issues in real time, shortening mean time to resolution and enabling responses before widespread impact. Further, machine learning recommendations increase in sophistication over time, as the system gains a better understanding of network and application behavior. So, over time, enterprises will be able to extend the autonomous functionality, by gradually loosening the conditions for manual review. 

Why should devops engineers care about a full-stack networking platform with machine learning and autonomous capabilities? Because it allows devops engineers to focus on the responsibilities they were hired for—building applications. When applications, services, or resources are added to the network or modified, the full-stack networking platform can see it and make necessary adjustments.

For example, if users are experiencing performance problems with a new application, the platform can quickly identify whether this is a routing issue or an application issue. Or if an application is delivered to users globally, the platform can ensure that all compliance requirements are met for each region in which it operates. And if the performance problem is related to the cloud region from which an application is delivered, the platform can determine if spinning up a cloud region closer to the users would improve performance and how much that would cost.

This points to an important benefit of a networking platform that works across clouds. For user-to-app connectivity, multicloud was shown to reduce network latency across 45% more paths than single cloud alone, while improving performance by up to 55% per path, compared to network-centric connectivity like VPN and SD-WAN, according to the State of Multi-cloud Infrastructure Report. Lastly, if security is most critical, the platform can prioritize security over performance for applications that traffic in sensitive data, or vice versa for latency-sensitive applications that don’t send and receive sensitive data. 

Another benefit of a full-stack networking platform is that it enables enterprises to align business and IT requirements while creating a simple, consistent, and scalable way to apply appropriate policies. This means devops engineers can quickly build, test, and deploy applications based on the needs of the business, instead of being limited to the timelines dictated by IT that we all experienced in the monolithic application era. A full-stack transit platform also overcomes any risks for the enterprise as a result of “cloud shadows,” the cloud counterpart to shadow IT, which experts anticipate will increase over the next five to 10 years. 

Further, with enterprises taking a more critical look at IT spend, including the growing percentage of cloud spend, a full-stack networking platform provides an easy way to understand if costs are in line and justified. All parts of the business can be required to show return on investment (ROI) or how they are contributing to revenue and growth. This approach helps devops engineers and line-of-business managers show ROI more easily as it relates to IT spend and innovation. 

Enterprises are struggling with IT complexity, with a shortage of cloud and security talent, and with the need to figure out how to overcome the friction between lines of business, devops, and IT teams. A full-stack networking platform—one built on cloud-native constructs, with multicloud support, enhanced with ML-driven autonomous capabilities that go beyond just networking or security—offers enterprises a way to overcome these challenges.

By eliminating friction and restrictions, and freeing devops to do devops, a full-stack networking platform with machine learning and autonomous capabilities sets up every part of the organization for long-term success.

Mehul Patel is head of marketing and customer insights at Prosimo.

New Tech Forum provides a venue to explore and discuss emerging enterprise technology in unprecedented depth and breadth. The selection is subjective, based on our pick of the technologies we believe to be important and of greatest interest to InfoWorld readers. InfoWorld does not accept marketing collateral for publication and reserves the right to edit all contributed content. Send all inquiries to newtechforum@infoworld.com.