The network is a crucial intersection for enterprises embracing a multi-cloud journey.
As enterprises migrate applications to cloud environments, their networks must provide reliable connectivity, support traffic flows and ensure secure access. But legacy network architecture often doesn’t adequately support applications in the cloud, according to industry experts discussing multi-cloud best practices at the ONUG Spring 2022 conference last week.
One challenge is that enterprises often try to apply their existing data center designs to new cloud strategies, said Neal Secher, vice president and head of network services at TD Bank. But that approach doesn’t translate to proper application performance.
“Don’t take legacy patterns, saying, ‘This is how I do it in non-cloud environments, so we’re going to conform it to cloud,'” Secher said.
Instead, enterprises should rethink how they marry applications and the network, determining what the applications need for performance, capacity and security — and which services the cloud providers natively support. Then, they can build down the stack to the network and avoid extending the legacy network to the applications, said Johna Till Johnson, CEO and founder of Nemertes Research.
“You need to structure the network to support the applications, not force the applications to behave in a network you built in 1995,” Johnson said.
Standardization vs. customization
A vital factor in multi-cloud strategies is the consideration of standardization versus customization when moving to a cloud environment. Each cloud provider offers its own set of differentiated features and services, and how enterprises host applications depends on design patterns that vary based on those features.
For example, some enterprises might opt to host an application with one cloud provider because it best matches that application’s use case, following a distributed deployment design pattern. Others might prefer to use redundant deployment patterns in which they deploy the same applications in multiple computing environments. The decision depends on use cases, provider capabilities and business requirements.
While enterprises can customize their multi-cloud designs, they should first standardize their architecture and deployment in a way that enables efficiency. As Shafeeq Shaikh, head of global network architecture, engineering and automation at Johnson & Johnson Healthcare, said, companies should standardize where they can and customize where they must.
Johna Till JohnsonCEO and founder, Nemertes Research
A company’s initial experience with its primary cloud provider can provide a foundation of the minimum capabilities required for efficiency, said Michael Wheeler, managing director of enterprise cloud and data center engineering at Cigna. Then, enterprises can use that foundation to determine the time and cost required to deploy applications in the next cloud, enabling a more standardized approach.
Secher agreed, recommending that enterprises opt into the native services that cloud providers offer, instead of trying to build and customize for each provider. While standardization might provide a lower set of capabilities, it enables enterprises to run their applications more quickly.
But standardization isn’t always as easy as expected. According to Arun Satyanarayana, senior distinguished engineer at Juniper Networks, it’s easier for developers to build a customized layer that runs across all cloud providers. Differences among providers make it harder to build a layer that’s standardized and repeatable for each cloud environment.
That’s why it’s so important for enterprises to assess each cloud provider, their network and security services, and available APIs, Shaikh said. While standardization can alleviate some of the sprawl and complexities that arise from varying design patterns, enterprises should take the time to create a clear roadmap.
“Multi-cloud is a journey, and you’ll be forced to make technical decisions,” Shaikh said. “Don’t make them until you have a clear strategy outlined; otherwise, you get inefficiencies in cost and governance.”