The market for enterprise server hardware has experienced a pretty boring run, as the basic server building block has not changed significantly in the last 20 years. However, the software that runs on top of servers is in the midst of major transition based on a new set of applications driven by Big Data, the Internet of Things, mobility and cloud-based services. The servers of yesteryear are no longer able to keep pace with the demands of this new generation of workloads where IT is now a focal point for competitive differentiation. As one approach to address this major shift in application requirements, companies like Cisco Systems, Hewlett-Packard Enterprise and Intel are bringing to market a new category of products called composable infrastructure.
In a recent report for my firm Moor Insights & Strategy, I outlined the market trends and key requirements for composable infrastructure solutions.
At a very high level, composable infrastructure is designed to treat hardware like software by allowing IT to manage “infrastructure as code”. Infrastructure that can be defined by software lets an IT administrator allocate the right amount of infrastructure resources (compute, storage, memory, I/O) needed to optimize application performance—no more, no less. Vendors promoting composable infrastructure solutions claim this approach will help IT lower total infrastructure costs, provide flexibility as resource needs change, and allow companies to get to market faster with their business-critical applications.
Enterprise IT is typically risk-averse when adopting new technologies, and composable infrastructure has the potential to increase IT complexity. With this in mind, it is important for vendors to make the transition to composable infrastructure as seamless as possible if they want to drive long-term market acceptance. I believe there are 4 key attributes IT organizations should consider when evaluating composable infrastructure solutions.
The Hierarchy for Composable Infrastructure
(Source: Moor Insights & Strategy)
- Disaggregated Programmable Infrastructure: Composable Infrastructure requires breaking apart the components of the system—compute, memory, I/O and storage—from each other so they can be allocated and managed as individual resources.
- Composition & Orchestration: Breaking apart a server is not very useful if you can’t put it back together again. An IT administrator must be able to compose and orchestrate the right components from the available resource pools to match the needs of each application.
- Automation & Management: A comprehensive management framework is required to automate discovery of resources in the infrastructure pool and implement policies to standardize the provisioning of those resources. This must include a set of APIs with broad support for 3rd party systems management tools to help ease integration into an existing datacenter environment.
- Application Centricity: The whole point of composable infrastructure is to match the right set of hardware to the needs of each application in order to maximize efficiencies. This can only be done if IT has a detailed understanding of what specific resources are needed to optimize performance for each of their applications and knows how to allocate the right amount of each resource without overprovisioning.
The market for composable infrastructure is still in its infancy. Leading enterprise server vendors are just beginning to bring their first generation composable infrastructure products to market. Also, it is yet to be seen how broadly and quickly the concept of software-defined infrastructure will be adopted by mainstream IT. I believe it will be critical for vendors to prove the long-term total cost of ownership (TCO) benefits of composable infrastructure solutions when compared to traditional datacenter infrastructure approaches to drive interest from IT and the potential for long-term adoption.
My full paper, The Journey to Composable Infrastructure is available here.