Shifting to Software-Defined
Showing results for 
Search instead for 
Did you mean: 

When it comes to composable infrastructure, beware the oxymorons


HP_PaulMiller_TH.pngBy Paul Miller, VP Marketing, HPE Converged Data Center Infrastructure

Lately, it seems, quite a few technology players are jumping on the Composable Infrastructure bandwagon, claiming that their take on a modern infrastructure is a truly composable one. I beg to differ. Upon inspection, many of these offerings appear to be piecemeal efforts dedicated to some part of the infrastructure, rather than the whole. To call them composable infrastructure, then, is a contradiction in terms – an oxymoron along the lines of an open secret or a definite maybe.

What, then, is Composable Infrastructure, you ask? To answer that question, let’s set aside fuzzy logic, and begin with what it’s not.

A server-based approach won’t cut it

yes.jpgComposable Infrastructure is not a server-based solution (see Paul Durzan’s post A Composable Infrastructure Bill of Rights). To call it so is to defy the very meaning of this new category of solutions. Rather, it’s an architectural approach that uses fluid pools of resources to dynamically configure applications using software and policy to optimize application and infrastructure performance. According to IDC, a Composable Infrastructure is all about “…looking at the IT assets at rack level or block level versus server level. Instead of thinking about individual servers and how each needs to be configured, IT can think at the rack level, or collection of racks level[i].”

A server that’s composable, as Cisco claims their M-series to be, can by definition only address a part of the infrastructure by pooling or sharing just the chassis and limited local storage across chassis compute nodes. According to Cisco, it’s applicable only to specific “cloud-scale, grid, electronic design automation (EDA), online gaming, and genomic applications[ii]”—a configuration that’s essentially useless.

A solution for all workloads … not just the privileged few

At HPE, we have a somewhat broader definition. For starters, a true Composable Infrastructure provides an open, industry-standard API that allows access to the entire infrastructure, one that supports all workloads—including databases, mission-critical, and SAN. Designed to fully disaggregate and converge compute, storage, fabric, and the operating environment, it works at an extremely high bandwidth to enable the kind of rapid composition needed for today’s digital economy, where infrastructure must be provisioned to create new services in a relative instant. The point of a true Composable Infrastructure is that it can be configured and reconfigured to meet the precise requirements of ALL workloads that must be run on it… not just a few. Those who would say otherwise are clearly confused.

A Composable Infrastructure is also not one that hampers innovators and creators with outdated technologies such as soon-to-be arcane Extensible Markup Language (XML), a set of “ridiculously verbose protocols that began with a commitment to simplicity and gave way to mind-numbing levels of complexity[iii].” The Cisco UCS Manager uses an XML API[iv], even though it’s common knowledge that leading cloud scale environments are moving away from XML to user-friendly JSON[v]. The mere suggestion is, well, seriously funny.

An infrastructure that’s a developer’s dream

On the contrary, a true Composable Infrastructure offers a modern, programmable interface within a thriving ecosystem, beginning with the API. An open RESTful API that uses JSON[vi], is now preferred by the vast majority of web-based developers, including those at Facebook and Twitter.

Finally, any forward-looking Composable Infrastructure program understands the importance of enabling DevOps for speed, innovation, and collaboration. ISVs and developers need easy programmatic access to the infrastructure to build applications, services, and other software required to win in the digital economy.

For this reason, a serious Composable Infrastructure needs to show partnerships with leading configuration management tools such as Ansible, Chef Software, Docker, and Puppet Labs. (Check out the HPE Composable Infrastructure partnership program.) Once you understand what Composable Infrastructure is, and what it can do, it’s easy to spot true leadership in the category. Look for specific features like:

  • A Docker Machine plugin that gives customers a way to automate the provisioning of physical infrastructure on-demand, from a private, bare-metal cloud using software templates that allow customers to treat infrastructure as code.
  • A Chef Provisioning Driver enabling organizations to order physical infrastructure on-demand from their private, bare-metal cloud using templates (or “recipes”) for unprecedented agility in the data center.
  • Partners such as Arista that can use open REST APIs and the open standard AMQP message bus to build true system interoperability. 

Composable Infrastructure, when delivered as a full design, is simply brilliant, providing the necessary building blocks for a revolution in enterprise data center infrastructure. Delivered incrementally, however, Composable Infrastructure amounts to little more than magic realism, and should be quickly side-stepped.


Keep up with HPE Converged Data Center Infrastructure on Facebook
Check out the new HPE Converged Systems Library
Follow us on Twitter @HPE_ConvergedDI
Join the Converged Infrastructure discussions on LinkedIn
Learn more about HP training classes at HP ExpertOne

[i] “Orchestrating Efficiency with Composable Infrastructure,” by Jed Scaramella, October 2015,


[iii] “Web services are dead -- long live REST” by Andrew C. Oliver,


[v] The hunt for the perfect API
New v2 API: What’s Not There Is Just As Important As What Is

[vi] HPE OneView Architectural Advantages

0 Kudos
About the Author