Covering Scientific & Technical AI | Sunday, December 22, 2024

Composable Infrastructure: Equal Treatment for Modern and Traditional Apps 

Most enterprises have a mix of modern apps that need adaptable infrastructure support and traditional software supported by old school IT. CIOs tend to solve this problem by focusing all internal development on traditional IT needs and writing checks to a public cloud provider to support the rest. But is giving up that type of control really the best solution?

A better model may finally be emerging. Composable infrastructure allows CIOs to program owned infrastructure as if it were a public cloud, allocating resources to serve both modern and traditional apps with equal deftness. Alabama's HudsonAlpha Institute for Biotechnology, a biotech research nonprofit, is embracing composable infrastructure for this very reason.

Composable Infrastructure: A Bridge over Troubled Waters

Rather than build infrastructure and then tune apps to fit what's in place, composable infrastructure is a pool of systems and services that can be arranged visually, from a high-level programming interface, using the tools that developers are already accustomed to.

As a non-profit genomics research institute with multiple missions—including incubating more than 30 life science companies-- one of HudsonAlpha's major tasks is to provide all its stakeholders tenants with the tools and resources needed to store, manage and manipulate vast amounts of genomic data, as well as handling the day to day needs of each of those stakeholders. Composable infrastructure accomplishes that by creating unique IT instances for the Institute’s 16 labs and all 30+ companies. All told, 200 employees on the non-profit side, as well as 600 employees of those for-profit associate companies, tap into some form of composable infrastructure to get work done at the HudsonAlpha genomics research facility in Huntsville, Alabama.

Composability allows for serving different needs easily. For example, researchers at HudsonAlpha handle approximately 15,000 genome sequencing projects each year. In digital form, each genome is represented by 100 to 300 gigabytes, which means storage needs add up fast. A hardwired SAN could serve the need but HudsonAlpha's team prefers the direct-attached approach for its speed and economics. With composability, researchers can activate DAS with a line of code, ordering it as it were an item on a menu.

The point here is that there's no such thing as a single-purpose enterprise. Every company has diverse IT requirements that can only be served by a set of diverse IT resources that interoperates seamlessly. Legacy applications must be tended to while new opportunities need new infrastructure, such as containers. Solving the problem by siloing traditional apps to on-premises infrastructure and outsourcing the rest to the public cloud creates friction.

How could it not? No corporate developer with a sense of pride wants to work on maintaining the SQL Server database when a cloud-driven mobile commerce project is the future of the company. Siloed infrastructure creates siloed teams, and siloed teams are a haven for resentment and poor performance.

We're All Cloud Giants Now

Composable infrastructure solves this problem by turning compute, storage, and fabric into a shared resource pool accessible as if it were any other code library. A high-level API does the work by acting as a bridge. Plug in the tool chain of your choice, develop, and then order the proper infrastructure mix at the time of deployment. Provisioning is done at the API level and then executed in the data center. Decoupling systems is just as easy.

Automation saves time and resources while sparing bruised egos. There are no "traditional" and "cloud" developers. There are just developers creating applications and assembling runtime environments -- both on-premises and in the cloud -- from a single, fully composable platform. Think of it as taking the cooperative mandate of the DevOps movement to new heights.

At HudsonAlpha, having composable infrastructure allows the institute's various teams to work in concert and glean value from more than one petabyte of new data every month. "We need to be able to integrate that data in real time and ask questions in real-time," said Jim Hudson, cofounder and chairman of HudsonAlpha.

He isn't alone. Rather, HudsonAlpha is one of a new and growing breed of entities built to bring ideas to market at unprecedented speed, and then capture the rewards of doing so.

Apps and data are the keys to thriving in this digital age, and good apps need good infrastructure. Until now, we've turned to the public cloud to fill that need. We don't have to anymore. With composable infrastructure, we can all be cloud giants.

Gary Thome is vice president and chief technologist for converged datacenter infrastructure at Hewlett Packard Enterprise.

AIwire