kantver - Fotolia


Understand the benefits, limits of composable infrastructure

Composable infrastructure leads the next stage of data center evolution, changing the way admins manage IT resources. But lock-in and other hurdles may hinder adoption -- for now.

While technologies such as virtualization, converged architecture and software-defined infrastructures accelerate the rate of workload provisioning and improve resource management, lingering silos and ongoing disconnects between hardware and software continue to restrain IT. Composable infrastructure -- a form of software-defined infrastructure that allows IT teams to scale resources and deliver them as a service in a matter of minutes or even seconds -- may be the next step in data center evolution.

Q. What is composable infrastructure?

Converged and hyper-converged infrastructures represent a major leap forward and show the value of hardware integration and unified management schemes. But convergence only extends to vendors' products, and rarely involves other resources or systems. There's still something missing.

The next major leap for data centers is the definition of infrastructure through software. Several iterations of this concept include software-defined storage and software-defined networking, but these ideas only deal with certain parts of the infrastructure. The ultimate goal is to consider all parts, together.

Rather than carve compute or storage instances out of certain systems, composable infrastructure uses software to discover and pool all data center resources, regardless of the resources' location or underlying hardware. It organizes those pools into service tiers and then uses highly integrated and automated tools to deliver compute, storage and network instances as services. In effect, administrators can "compose" infrastructure on-demand from resource pools to provide the required level of service for a workload. Administrators no longer need to worry about which servers, disks or networks are involved.

Composable infrastructure overlaps several other recent IT concepts, including infrastructure as code (IaC), software-defined infrastructure, software-defined data centers (SDDCs), disaggregation and software-defined everything.

Q. How does composable infrastructure differ from virtualization and software-defined technology?

The two technologies are related, but composable infrastructure is a more recent concept that builds on the underlying capabilities of virtualization. Virtualization abstracts enterprise workloads from the hardware resources they run on. This allows IT teams to eliminate resource waste in servers, and reduce the total number of physical servers in their data center. Virtualization also allows admins to migrate workloads between local and remote machines at-will, rather than install and tie them to specific physical machines.

But despite these benefits, IT admins still have to manually provision virtual machines (VMs). And this presented the need for an even greater layer of abstraction -- which is where software-defined technologies come into play.

Software-defined technologies have evolved in several key areas, including software-defined storage and software-defined networking. Later, the notion was expanded even further to embrace an SDDC. These technologies are all based on a software framework that uses virtualization to abstract logical entities -- such as VMs or virtual network segments -- from the underlying physical hardware.

More recently, software-defined infrastructure underwent another conceptual change that recognizes the importance of code to drive a software-defined environment -- or composing resources to service workloads. Consequently, software-defined is also referred to as IaC and, most recently, composable infrastructure.

Q. Are there any composable infrastructure products?

There are several vendors with composable infrastructure offerings, including Cisco and Hewlett Packard Enterprise (HPE).

Cisco had originally touted its M-Series Unified Computing System (UCS) for composable infrastructure deployments, but the M-Series has since been discontinued and Cisco points customers to UCS C-Series rack servers and UCS B-Series blade servers. Both iterations are available in various models.

HPE focuses on composable infrastructure with its Synergy platform. HPE Synergy provides compute, storage and fabric as disaggregated resource pools that you can instantly configure through software to meet the specific needs of each workload. Since composable infrastructure depends on software for provisioning, automation and intelligence, there is a strong emphasis on APIs to integrate the platform with third-party management, reporting, automation and other tools.

Other vendors like Dell haven't been as quick to embrace the notion of composable infrastructure, instead focusing on alternative modular approaches designed to abstract physical infrastructure. The ultimate fate of composable infrastructure -- whether it continues as an IT concept or morphs into some other technology -- depends on the amount of traction these products receive in the marketplace.

Q. Are there any limitations to composable infrastructure?

The biggest limitation to composable infrastructure right now is vendor lock-in. Only a tiny number of vendors, including HPE and Cisco, are actively in the marketplace. Given the lack of significant standards, each vendor approaches the technology with its own slant -- which tends to make resulting products proprietary and monolithic.

But composable infrastructure requires openness and interoperability. After all, the point is to automatically pool and manage any resources from any vendor anywhere in the data center. Composable infrastructure should work just as well for acres of white box servers and multivendor disk arrays as it does for a single vendor's offering.

The software stack is another concern. Composable infrastructure uses APIs to access and manage resources. The APIs that are central to software-driven resource provisioning and management must be simple, open and extensible so that any developer can create code to drive the infrastructure. The resulting code base should work the same way on any composable infrastructure platform located anywhere.

Admins must vet and carefully test other software components involved in the overall stack, such as virtualization, automation and orchestration. Any software flaws, integration issues or poorly implemented upgrades can have a profound effect on the overall behavior and reliability of composable infrastructure.

There is some movement in the software area. Standards such as DMTF's Redfish API are being extended to address all of the components in a data center using a single, consistent API to discover, inventory, configure, organize, orchestrate and monitor components within the infrastructure. Other software platforms like OpenStack find use for software-defined functionality like orchestration and automation. Still, the broader industry won't agree on or adopt software standards for common operation of composable infrastructures anytime soon.

Next Steps

Using converged and disaggregated infrastructure side by side

HPE Synergy points to possible composable integration

A full SDDC presents new IT hurdles

Dig Deeper on Virtualization and private cloud