The power of purpose-built cloud infrastructure [MIT Tech Review]

View Article on MIT Tech Review
While AI is accelerating cloud adoption, organizations’ reasons for migrating their systems and applications to the cloud remain relatively consistent: a desire to lower capital expenditures, increase agility in a fast-paced business environment, and improve availability of business-critical resources.

Flexera’s 2024 State of the Cloud Report underscores organizations’ consistent desire to make the most of cloud: Asked about their cloud initiatives for the year, 71% of respondents reported that they were working to optimize their existing use of cloud—making this the top cloud initiative for the eighth consecutive year. Organizations are also still working to shift more of the business to cloud: The next-most-popular initiatives were migrating more workloads to cloud (cited by 58%) and progressing on a cloud-first strategy (48%).

But while moving to the cloud will always deliver significant benefits, the demands of business are fast evolving, requiring organizations from all industries to support a new level of compute power. Today’s banks must build models capable of measuring and monitoring credit risks. Manufacturers must rapidly iterate on product designs to reduce time to market without compromising quality. And pharmaceutical companies require high-performance infrastructure to accelerate insights in genomics and precision medicine.

In these specialized scenarios, running all workloads on the same general-purpose infrastructure is no longer sufficient. Organizations now have a wide variety of cloud options, and they need to make infrastructure choices that best fit their needs and use cases. Making uniform choices has consequences for cost, performance, and scalability, according to Paul Nash, corporate vice president of product for Azure Core Infrastructure at Microsoft. “General purpose really isn’t general anymore, so understanding how to match the right products to your solutions becomes a significant part of capturing the value that cloud offers,” he says.

Because of this, organizations also cannot lift and shift existing systems to the cloud without working through these decisions. “If you try to move an on-premises architecture into the cloud without really thinking through the requirements,” says Nash, “you won’t achieve optimal performance from the cloud.” As they migrate, he adds, companies must factor in what cloud changes: “among other things, the network is different, the disaster recovery plans are different, and the data storage mechanisms are different.”

Especially as modern cloud-native workloads running on Linux and open-source software multiply, purpose-built cloud infrastructure is emerging as a better approach for leveraging the benefits of cloud while, at the same time, ensuring the flexibility and agility needed to grow and evolve in today’s modern dynamic environment. 

The power of purpose-built infrastructure

In the past, cloud environments were highly standardized and modular: servers had the same rack width, power configurations, and storage, regardless of application needs.

That’s changing with the rise of purpose-built cloud infrastructure, which consists of a complete set of computing, networking, and storage resources fully integrated with workload orchestration services for high-performance cloud applications. Purpose-built cloud can be customized to handle specific workloads, ensuring low latency, high throughput, and greater resource utilization. Security measures and compliance measures can be easily integrated to satisfy the standards of highly regulated industries such as health care and finance. And built-in features such as redundancy and disaster recovery capabilities help achieve high availability of business-critical applications.

As a result, purpose-built cloud infrastructure enables access to vast compute resources from a cloud platform that not only provides scalability, flexibility, and performance but is also able to support the growing complexity of modern cloud-native workloads.

“At the end of the day, everything is about performance, security, and power,” says David Harmon, director of software engineering for AMD. “With a purpose-built cloud infrastructure, organizations can do just as much with the same compute power at a lower total cost of ownership while still maintaining performance.”

AI innovation driven by cloud

As executives recognize its potential to create innovative applications, from life-saving medical devices to self-driving cars, AI is driving demand for cloud infrastructure. But supporting AI’s high demands for compute power, storage, and networking capabilities requires purpose-built AI supercomputers in the cloud.

In fact, according to a Forrester Consulting survey, 91% of respondents say AI infrastructure is critical for business success. Yet more than half—56%—say their organizations don’t have the proper infrastructure to support desired AI workloads.

Fortunately, the right cloud infrastructure can prevent organizations from having to start from scratch when enabling their applications for AI. The models and services cloud providers offer give organizations a running start: “They’re not starting from zero or building from the ground up,” says Nash. “They’re leveraging the technology built into a purpose-built infrastructure to innovate faster for their business.”

Cloud also enables businesses to keep pace with AI innovation. “AI was built knowing that it was going to run in cloud,” says Nash. “And what that means is that it can innovate and evolve at the pace that cloud moves.” If a manufacturer releases a new GPU (graphics processing unit), for example, it can be accessed in the cloud the next day. “Organizations can access the latest and greatest technology as soon as it’s available. They don’t have to wait for procurement or sign up with a new vendor,” says Nash.

Purpose-built infrastructure for AI workloads in the cloud is also paving the way for new innovations across industries. Retailers, for example, can accelerate the training of their machine learning algorithms to react in real-time to evolving customer demands and emerging market trends. Manufacturers can take advantage of unprecedented amounts of sensor and operational data to proactively predict equipment failures and avoid costly downtime. And health-care organizations can accurately forecast patient demand, reducing the risk of out-of-stock medications or other life-saving deliverables.  

Putting best practices in place

Delivering value from purpose-built cloud infrastructure requires more than fully integrated computing, networking, and storage resources. Organizations also need a thoughtful approach to how they plan for and deploy cloud infrastructure.

Nash recommends that organizations start by carefully considering what they want to accomplish with their software architecture. By examining their business objectives, organizations can pinpoint which of their workloads can most benefit from purpose-built cloud infrastructure, and then plan their cloud deployment accordingly.

Once an organization determines how to reap the greatest return on its cloud infrastructure investment, Nash says it must then be prepared to shift its teams’ focus from labor-intensive tasks such as infrastructure management, services management, and backup, to more business-critical activities, such as product innovation and customer engagement. “The more time an organization can spend focusing on innovation, the more value it can create for the business,” says Nash.

When it comes to taking advantage of the latest advances in AI and Linux-based and open-source cloud-native services, security must also be top of mind. For this reason, Nash recommends that organizations work with cloud providers like Microsoft Azure that have early access to information on risks that might impact firmware and operating systems. IT teams should also leverage an infrastructure’s built-in management capabilities to handle security updates and consistently monitor their security posture.

Most importantly, a purpose-built cloud infrastructure must support a positive customer experience. Nash says that organizations must consider how their applications are architected and where critical connections exist to avoid negative user experiences, such as high latency. Deploying workloads in different geographies, using multiple points of presence, can help ensure that workloads in the cloud deliver the best possible customer experience.

A platform for today and tomorrow

Today’s organizations are increasingly moving to the cloud to maximize return on investment, maintain business performance, stay competitive, and capitalize on AI. But while migrating to the cloud has provided IT leaders with flexibility and simplicity in how they manage their resources, it has also caused a rapid increase in complexity.

Fortunately, purpose-built cloud infrastructure can help by maximizing the performance of resource-intense workloads, meeting the highest security and compliance standards, and ensuring high availability of critical resources. And in today’s race to build and train advanced AI applications, purpose-built cloud infrastructure empowers developers to deploy high-quality models faster, boosting their potential for AI breakthroughs.

Maximize ROI and performance for your Linux and open-source workloads with Microsoft Azure and AMD. Learn more about Linux on Azure.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.