WavebreakmediaMicro - Fotolia

Composable infrastructure platform FAQs get answers

While the goal of hyper-converged infrastructure and composable infrastructure are much the same, the way they achieve that objective of on-demand IT infrastructure differs.

A composable infrastructure platform brings the ease of management of hyper-converged infrastructure to IT and data center admins without the rigidity of lock-step scaling of all the components. However, this achievement often requires more complex coding and can result in tighter vendor lock-in.

The goal of composable infrastructure is to finally achieve a software-defined data center, where any physical server, virtual machine (VM) or even container can be deployed out of a shared pool of resources of compute, storage and network interconnections. To do this, a composable infrastructure platform disaggregates the resources in the physical servers of a data center. It uses software to make compute, storage and network connectivity available out of shared pools for deploying on-demand servers, VMs or even containers.

Unlike hyper-converged infrastructure (HCI), which uses hypervisors to virtualize and manage resources, composable infrastructure systems use a unifying API to manage disaggregation and manage the shared pool of IT resources. A major advantage of a unifying API is it allows the creation of preprogrammed scripts or templates that make deploying infrastructure on demand as easy as running the script -- if the infrastructure you need is already defined as one of those templates. This makes composable infrastructure operate much like Ansible or Chef, or containers like Docker Composer.

Because the network interconnect resources are shared in a composable infrastructure platform, a couple of companies have introduced products that share across the fast PCI Express (PCIe) bus. Kaminario recently introduced Kaminario Flex software into its all-flash arrays based on nonvolatile memory express, which enables them to be shared as part of a composable infrastructure. Startup Liqid has introduced a PCIe fabric switch for composable infrastructure that enables any component attached to a server via a PCIe connection to be shared. That includes GPUs for extra-powerful compute resources.

What is disaggregation, the concept at the core of both a composable infrastructure platform and HCI?

To understand disaggregation, you have to realize that a standard data center server is an aggregation of various devices in one box or blade. The processor and RAM provide compute power. The HDDs or SSDs provide data storage. And the network cards and fabric provide interconnection to other servers.

Disaggregation in either composable infrastructure or HCI uses a software layer that determines how much capacity each of those three resources has available from that single server and adds that capacity to a shared pool of resources that can be used by any application or workload as needed.

Who first used the term composable infrastructure?

While it is unclear where the term composable infrastructure originally came from, it first became common as the term used by Hewlett Packard Enterprise to describe its Synergy product.

What are the differences between composable infrastructure and HCI?

The two main differences between a composable infrastructure platform and an HCI platform are how IT administrators can scale them and how they manage shared resources. HCI systems use a hypervisor to manage all of the virtualized resources that have been combined into the shared pools. Composable infrastructure uses a unifying API to manage the shared resources.

When it comes to scaling the capacity of the data center, HCI -- particularly HCI appliances -- have to be scaled as nodes containing new compute, storage and network capacity. Individual components can't be scaled by themselves. With composable infrastructure, physical components such as storage can scale individually without having to scale all components in the physical server.

What is a composer?

The word composer can define many things, including one who writes music, but in a composable infrastructure platform, it refers to the software that "composes" an aggregation of resources for an application or workload on demand. The composer can place that application on a bare-metal server, in a VM or in a container, depending on templates that have been defined by IT admins based on the location and amount of shared resources that work best for a specific application.

Is a composable infrastructure the same as a data center operating system?

Data center operating system, or DC/OS, is primarily a marketing term used by companies like Mesosphere to describe its composable infrastructure. Some experts have determined there is no such thing as an actual DC/OS, because even fully featured composable platforms like Mesosphere or HPE Synergy can't manage core aspects of the data center, such as cooling and power management.

How is composable infrastructure related to infrastructure as code?

Infrastructure as code, or IAC, uses preprogrammed scripts to automate the deployment of infrastructure on demand. There are stand-alone products that can do this, including Ansible and Chef. Many vendors of composable infrastructure use the same concept to simplify the deployment of infrastructure built from the pooled resources available for the IT admins. The composer software in a composable infrastructure platform functions as an IAC system.

What are the downsides to composable infrastructure?

While a composable infrastructure will increase the flexibility of scaling, it has the same vendor lock-in problem as with most HCI systems. You are restricted to the vendor's physical servers and devices, or a list of compatible hardware the vendor has verified. And while using preconfigured scripts or templates that define the infrastructure to be deployed make IT management easier, you need specialized skills to set up those templates and adjust the code.

Dig Deeper on Converged infrastructure management

Cloud Computing
and ESG