Data centre infrastructure has evolved quickly in recent years, with converged and more recently hyperconverged systems offering greater control through a software layer.
According to HPE the latest step in this evolution towards a fully 'software-defined' data centre is composable infrastructure, namely its Synergy platform, announced last year.
The idea behind composable systems is that hardware components – compute, storage and networking – are managed entirely via software commands. Pools of resources are created and assigned automatically to meet the demands of individual applications in near real-time.
This contrasts with the siloed approach of traditional IT architectures.
"We are now at the point with composable infrastructure where you have infrastructure as code: literally a single command line or line of code to deploy a piece of resource to allow you to run your infrastructure," says David Chalmers, HPE's vice president and chief technologist, EG EMEA.
Analysts believe that there will be a lot more action from the vendor community in the next year. A recent research note from Moor Insights and Strategy states: "Over the next 12 to 24 months, the market is expected to ramp with additional new products, tighter integrations across vendors, and usability enhancements to make deploying and managing composable infrastructure easier for mainstream IT organisations to adopt more widely."
Composable infrastructure: What are the benefits?
There are a number of potential advantages, according to Chalmers. Greater utilisation of resources is likely to result in cost savings with regards to capital expenditure.
It can also speed up the deployment of applications, helping to meet demands from the wider business to create new products.
In addition, composable systems allow IT teams to bridge the gap between older systems and newer cloud-native applications.
"We see Synergy as being the way to give the on-premise customer the flexibility of use that they need," says HPE's Chalmers. "It is not just buying one style of system for ERP and a different set of systems to run CRM, and then something else for different application modes."
"When you do that in the converged or even hyperconverged world, you end up with a very complicated mixture of systems. Composable is about saying 'don't do that, put them in a physical infrastructure that you can change very quickly'.
"So when ERP systems need [resources] because it is the month end that is fine, we move the resource around. If it doesn't need more because it is the first week or the quarter we divert that to the marketing application and so on and so forth."
All of this, says Chalmers, provides a compelling alternative to moving workloads out to the public cloud. Despite the claims of public cloud vendors, many businesses will want to keep at least some of their systems on-premise, particularly for regulatory purposes.
"People are recognising that being focused on any one version of the approach – it is all public or all private - is not the way to go. Composable infrastructure gives people a dramatically more effective option on-prem to get that flexibility and control they need."
Composable infrastructure: 'Hyperconvergence' rebranded?
The 'composable' concept bears a resemblance to hyperconverged systems that have brought improved software management to hardware systems.
However, there are some key differences. In contrast, says Chalmers, hyperconverged systems are more suited to single workloads that require scalability.
"We see a lot of people talk about 'is this just a fancy way of describing hyperconverged', and we would say no," says Chalmers.
"Hyperconverged is a step on the way. It tends to be where you have the same style of application running on all of the nodes. Because the nodes are relatively fixed in size, the whole point about hyperconverged is that it is incredibly simple, but it is also not designed to be flexible. It is designed to allow you to run a similar workload in a very easy way."
Moor Insights & Strategy's report suggests that organisations already using hyperconverged systems are "a good target for a transition to composable infrastructure".
The note states: "These customers are already familiar with the benefits of a single platform with server, storage, and network resources – and the efficiencies from an application-centric approach to resource deployment.
"Organisations that decide to take the next step to composable infrastructure from a converged or hyper-converged platform could experience additional benefits by pooling these resources and dynamically provisioning / reprovisioning their infrastructure as workload needs evolve."
Composable infrastructure: A step towards true software defined data centre
Looking at the bigger picture, Chalmers sees composable infrastructure as another stepping-stone in the evolution of server architectures to the memory-centric - as opposed to processor-driven - architecture exhibited in its 'The Machine' project the long term goal.
"We had mainframes and we had minis and we had client servers, then we had server-client, internet and converged, now we have composable. So what comes next?," Chalmers says.
"The answer is 'mesh' and memory centric computing that is built around fundamental architectures turned round to be memory-centric, with different pools of resource.
"But it is built upon the composable concept of fluid pools of resource: instead of having processors tied tightly to memory, you separate memory from it."