Composable Infrastructure Advances SDDC Dreams
Park Place: Hardware Maintenance
HCI has taken IT a long way, but it still relies on pre-configured systems, and siloes and exceptions remain. Additionally, scalability is generally limited to about 25 nodes. In other words, there was room for improvement.
The software-defined data center, or SDDC, is the nirvana most enterprise IT pros are shooting for in their private cloud slash on-premises infrastructure. As we’ve pointed out before, there is a long, winding road to SDDC, but the industry has added an important new station.
Composable infrastructure is taking off from the whiteboard to the product line-up. The goal is truly virtualized, highly flexible infrastructure, or as some like to say “infrastructure as code.” It’s the next stop after hyperconverged and offers an array of benefits.
Traditional to Composable, the Path Most Will Travel
We’re all familiar with traditional infrastructure. It’s the basis on which data centers were once built. From there, converged infrastructure (CI) developed, leading to hyperconverged infrastructure (HCI), and now the true target destination in composable infrastructure. Here’s how the various stages differ:
- Traditional infrastructure: Compute, storage, and networking are separate.
- Converged Infrastructure (CI): Compute, storage, and network become physically integrated. Hardware and software comes preconfigured but systems are only applicable to a particular workload or “silo.”
- Hyperconverged Infrastructure (HCI): A hypervisor-controlled “all in one” CI. Won’t accommodate physical and SAN-attached applications.
- Composable Infrastructure: Customer reconfigurable infrastructure with Compute + Storage + Networking on a single platform
As you can see, HCI has taken IT a long way, but it still relies on pre-configured systems, and siloes and exceptions remain. Additionally, scalability is generally limited to about 25 nodes. In other words, there was room for improvement.
Composable infrastructure is a significant technological leap, and like other advances before it, composable infrastructure can be combined with other approaches, notably HCI. But what is it good for anyway?
Why “Compose” Infrastructure?
Composable infrastructure is extremely public cloud-esque in its advantages. Compute, storage, and network all become resource pools to be provisioned on the fly as workloads demand.
That means a developer can specify what capacity of each is needed for a workload, “borrow” that precise capacity from the pool, and then “return it” when the workload is finished. If one workload is compute-heavy and the next is memory-heavy, provisioning will follow suit. This is different than running dedicated VMs, as in a hyperconverged environment.
In other words, composable infrastructure offers the provisioning flexibility of the public cloud (IaaS), but the systems remain on the client’s premises, adding all the control, security, and compliance advantages of in-house IT.
The dynamic resource sharing enabled by composable infrastructure can substantially reduce asset underutilization, which plagues the enterprise data center. In most cases, utilization rates top out at about 35% but often hover between 20% and 30%. This is life where resources are fixed.
Fully optimized public cloud utilization is closer to 95%. Although that’s not all about architecture, composable infrastructure can help enterprises reach far higher utilization rates. Back-of-napkin estimates say twice the current utilization rate is within reach, so there is a very promising return on investment.
Another feature of composability, it isn’t limited to a single technology. It’s fine with VMs, containers, bare metal, and cloud-native applications. It can run anything. It can store anything.
Composability will reduce enterprise IT complexity, lower costs, support continuous delivery (CD), and aid in the constant quest for more uptime. Finally, it will deliver advantages as computing moves to the edge, but we’ll get to that later.
Who’s Offering Composable Infrastructure Products?
HPE is considered a leader in composable infrastructure with its Synergy line. Intel’s Rack Scale Design, which we covered in greater depth here and here, is a form of composable infrastructure as well. Moreover, Cisco is treating infrastructure as code with Intersight, Dell EMC is marketing its kinetic infrastructure, Nutanix is jumping from HCI into composable, and startup Liqid is tackling the composable problem, including the all-important PCIe switch, from the ground up.
None of these OEMs are able to make the data center fully composable at this time. Eventually, APIs will request resources without any IT humans involved, and the data center will be self-managing with dynamic provisioning of virtual servers optimized to the performance needs of each application and workload.
Composable infrastructure is a heady concept and 2018 looks like it might be the year it really starts to take hold.