Victoria - Fotolia

Composable infrastructure lets applications take the lead

A growing data center architecture trend known as disaggregation or composable infrastructure has a new entrant with HPE Synergy.

Traditional data center architecture groups resources together, but the growing concept of composable infrastructure looks to separate each component while maintaining the same workflow.

It is known as disaggregation -- or composable infrastructure -- and it involves breaking apart the various components of the data center IT infrastructure and bringing them together in new ways to meet the needs of specific workloads.

This idea of infrastructure as code means that resources are looked at as individual assets -- fabric, compute and storage -- with the ability for storage, for example, to be pulled from another server, removing that as a constraint in the data center.

"Now you can look at it from a rack level as a shared resource pool," said Jed Scaramella, research director at IDC Research Inc. in Framingham, Mass.

This week, Hewlett Packard Enterprise (HPE) unveiled its first hardware associated with the composable infrastructure initiative, Synergy. HPE Synergy will be available in the second quarter of 2016 and pricing information has not yet been released.

How composable infrastructure works

Synergy makes available "fluid resource pools" of compute, storage and networking that do not need to be on same server and work across physical and virtual environments as well as containers, said Paul Durzan, vice president of product management for HPE's converged data center infrastructure.

Synergy's composer defines the footprint of the application by drawing the exact resources needed from the resource pool. The composer also automatically recognizes a resource when it is plugged in and elevates it above the form factor.

Instead of putting an application on a physical server that may not have exactly the right resources, IT professionals can pick out specific resources with composable infrastructure.

"It will give that application the exact set of resources it needs," Durzan said. "You can start to span your data center at a higher level."

When complete, HPE Synergy will uninstall the template and the resources will go back into the pool and become available for the next application. The templates created for a specific application using Synergy's software-defined intelligence allow IT to optimize the infrastructure for each application.

Disaggregation also helps reduce the number of zombie servers in a data center by putting them to use for applications.

"What you are seeing is a vision of taking a physical infrastructure and letting it span the physical boundaries," Durzan said. "This helps reduce the number of stranded resources in the data center."

While composable infrastructure is a new category, it is based on x86 server architecture and has a disk drive that can use spinning or hard disk memory.

"The technology underneath has all been there," Scaramella said.

Disaggregation limitations

HPE Synergy allows the CPU to be decoupled from the storage and memory and allows infrastructure to become compute- or storage-intensive. It is the first generation, though, and certain components remain tied together; CPU and memory need to remain close together because of latency. Silicon photonics will allow the two to be further apart in the future, he said.

The biggest question for prospective HPE Synergy customers is how easy it will be to decouple the assets -- and they will want to see it work.

"They need to gauge that the [service level agreement] will be met," he said.

HPE could have sold its intelligence software separately with all the pieces of composable infrastructure together without the Synergy hardware, but HPE has "smartly packaged it" with hardware, Scaramella said, adding that selling the hardware provides "technology that is suited for composability."

It brings cloud-like flexibilities to the on-premises layer.
Richard Ficheravice president and principal analyst, Forrester Research

There's nothing similar on the market, according to several analysts, although the closest is the M-Series from Cisco. The M-Series is aimed more at service providers with dynamic workloads than the enterprise customers Synergy aims for, Scaramella said.

Cisco UCS M-Series Modular Servers were the first commercially available disaggregated system. The physical disk on the M-Series is hidden and it can be portioned to 16 servers, but it is not as rich as Synergy, according to Richard Fichera, vice president and principal analyst at Forrester Research Inc. in Cambridge, Mass. The HPE c-Class, Cisco UCS and Dell PowerEdge FX are all similar, he said.

HPE's newest system portends more to come in the composable infrastructure market, Fichera said.

"You don't have to be a rocket scientist to think that others are looking at this," Fichera said. "The HPE offering is pretty impressive; it is going to blow things open with people talking about composable."

Composable infrastructure mirrors cloud flexibility

Two of the most significant benefits will be shortened provisioning time and increased flexibility, he said.

The on-premises flexibility from Synergy will be similar to infrastructure as a service providers that offer bare metal servers such as Internap Corp. and IBM's SoftLayer Technologies Inc.

"It brings cloud-like flexibilities to the on-premises layer," he said.

It also puts a cloud-like operational framework in an enterprise data center with software control and templates that can quickly be deployed.

Composer can find and discover all available resources and can also add additional frames and auto-discover and auto-assimilate up to 20 frames from single management endpoint.

"This is the physical realization of composable infrastructure -- the architecture itself," said Gary Thome, vice president of blade system strategy at HPE and chief engineer for Synergy.

Synergy appears to be the next generation of HPE's blade servers and though previous generation blades will be available for the foreseeable future, this new approach to blades will ultimately replace the older hardware, Scaramella said.

HPE Synergy automates through built-in APIs

Synergy also includes a unified application program interface (API), which helps further automate the data center. It can take up to 120 hours to automate the typical application, Durzan said, and a unified API across infrastructure makes it simpler, with one line of code that can call all parts of the infrastructure and eliminate many of the manual tasks required to automate the data center.

"We can now integrate into a large ecosystem," Durzan said, including partners Ansible Inc., Chef Software Inc., Puppet Labs and VMware Inc.'s vCenter with one API that can program the entire infrastructure.

Synergy is the latest in what analyst firm Gartner Inc. calls "fabric-based computing," said Paul Delory, a research director at Gartner in Stamford, Conn.

"This is not a new idea," Delory said. But, "this is the closest anyone has come so far."

IT professionals often talk about being able to have a CPU sled and RAM sled "and be able to carve those up any way they want," he said. The limiting factor is the Intel Xeon architecture, and that level of disaggregation is not possible with Synergy, he said.

Because of that, the full potential of a composable infrastructure has not been realized, Delory said. "I think they have a vision and the Synergy product is not the full realization of the vision," he added.

Next Steps

Server architectures are about to get interesting

AMD looks to future in data centers

Intel's approach for scaling up

Dig Deeper on Server hardware strategy