denisismagilov - Fotolia
Composable architecture is a trend in enterprise data centers that brings programmatic control to server hardware deployment and configuration. Vendors claim that with composable architecture -- also known as composable infrastructure -- you can have the virtualization benefits of a hyper-converged infrastructure while being able to expand individual resource components separately as needed, like traditional IT architecture.
The idea of composable architecture is that you maintain a pool of physical resources and assemble servers or workstations through software to run a workload, much the same way hyper-converged infrastructure (HCI) operates. The major difference between composable infrastructure and HCI is that HCI has all the physical resources in a set of single appliances or hardware nodes. A composable infrastructure system can use separate devices for compute, storage and networking.
Hewlett Packard Enterprise said that it invented the term composable infrastructure to describe the capabilities of the Synergy platform, which combines conventional blade servers with software-defined storage and network connectivity, plus an operating system and application deployment system.
In composable infrastructure, servers can be virtual, containerized for DevOps purposes or even physical servers that still get their compute, storage and fabric resources delivered from the virtual resource pool. IT administrators can often assemble these servers on demand based on the need to test or deploy an application. Part of this software-controlled deployment process for an application would be to assemble the physical computer and then implement the required operating system and application. In the case of physical servers in a composable infrastructure, this is bringing the agility that virtualization gives to VMs into bare-metal application delivery.
Composable hypervisor host
One of the use cases for composable architecture is deploying a virtualization cluster to test an application, then destroying it once the test is complete and returning the resources to the pool. If the application testing is completed on composable infrastructure, then composable infrastructure should also be used for production deployment. That guarantees the highest fidelity between test and production. Once the virtualization platform is on composable infrastructure, there are a few points to consider, mostly positive:
Consistency. Composable infrastructure allows admins to build multiple servers with the same configuration, using the same profile to deploy a whole cluster of virtualization hosts. This consistency with hardware nodes is the foundation of a consistent virtualization cluster build.
Scaling out and in. Virtualization clusters often need to grow and occasionally shrink, as the workload on the cluster changes. Using the same profile, new hardware nodes can be deployed the same as the existing servers. Equally, hardware nodes can be removed from a cluster and resources returned to the pool for re-use.
Scaling up and down. The CPU and RAM available in a single hypervisor host is a fundamental limit to the size of the virtual machine (VM) that the host can run. As we see more in-memory compute, and CPU-heavy applications such as machine learning, there is a potential benefit to being able to increase the size of the virtualization host available for a workload. Composable architecture simplifies the process to deploy larger hosts, and some versions plan to also allow physical servers to grow and shrink CPU and RAM on demand.
Generational change. Over time, new servers with new CPU types and even CPUs from different vendors can be purchased and added to the composable architecture platform. The software control makes it simpler to deploy new clusters and migrate older hardware nodes out. By integrating with software control of the VMs, it is possible to automate the migrations and minimize the outage for individual VMs and applications.