There’s an evolution occurring in IT infrastructure that’s providing alternatives to the traditional server-, storage- and SAN-based systems enterprises have used for the past two decades or so. This evolution was first defined by “hyperscalers”, the big public cloud and social media companies, and encompasses multiple technology approaches like Converged, Hyperconveged and now Composable Infrastructures. This blog will discuss these combined “Integrated Infrastructure” approaches and look at how they attempt to address evolving needs of IT organizations.
Companies like Google, Facebook and AWS had to address a number of challenges including huge data sets, unpredictable capacity requirements and dynamic business environments (first, public cloud and social media, now IoT, AI, etc.) that stressed the ability of traditional IT to deliver services in a timely fashion. So they created a new model for IT that incorporated software-based architectures (developed internally), that ran on standard hardware and provided the needed flexibility, scale and cost-containment.
But enterprise IT doesn’t have the expertise to support this kind of do-it-yourself infrastructure, nor the desire to dedicate the resources or take on the risk. In general, enterprises want to use trusted suppliers and have clear systems responsibility. They need much of what hyper-scale systems provided, but with integrated solutions that are simple to operate, quick to deploy and easy to configure and re-configure.
CI, SDS and HCI
Converged Infrastructure solutions were some of the first attempts at an integrated infrastructure, creating certified stacks of existing servers, storage, networking components and server virtualization that companies bought by the rack. Some were sold as turnkey solutions by the manufacturer and others were sold as reference architectures that VARs or enterprises themselves could implement. They reduced the integration required and gave companies a rack-scale architecture that minimized set up costs and deployment time.
Hyperconverged Infrastructures (HCIs) took this to the next level, actually combining the storage and compute functions into modules that users could deploy themselves. Scaling was easy too, just add more nodes. At the heart of this technology was a software-defined storage (SDS) layer that virtualized the physical storage on each node and presented it to a hypervisor that ran on each node as well, to support workloads and usually the SDS package itself.
HCIs come in several formats, from a turnkey appliance sold by the HCI manufacturer to a software-only model where the customer chooses their hardware vendor. Some enterprises even put together their own HCI-like solution, running an SDS package on a compatible server chassis and adding the hypervisor.
While Converged and Hyperconverged Infrastructures provide value to the enterprise, they don’t really provide solution for every use case. HCIs were great as a consolidation play for the lower end, SMB and mid-market companies. Enterprises use them too, but more for independent projects or remote environments that need a turnkey infrastructure solution. In general, they’re not using HCIs for mainstream data center applications because of concerns about creating silos of infrastructure and vendor lock-in, but also a feeling that the the technology lacks maturity and isn’t “mission critical” (based on 2017 Evaluator Group Study “HCI in the Enterprise”).
While they’re comprised of traditional IT infrastructure components, CIs present a system that’s certainly mature and capable of handling mission critical workloads. But CIs are also relatively expensive and inflexible, since they’re essentially bundles of legacy servers, storage and networking gear, instead of software-defined modules of commodity hardware with a common management platform. They also lack the APIs and programmable aspects that can support automation, agility and cloud connectivity.
Composable Infrastructure (CPI) is a comprehensive, rack-scale compute solution that combines some characteristics of both Converged and Hyperconverged Infrastructures. CPI disaggregates and then pools physical resources, allocating them at run time for a specific compute job, then returns them to the pool. It provides a comprehensive compute environment that supports applications running in VMs, containers and bare metal OS.
CPI doesn’t use SDS, as HCIs do, to share the storage pool, but supports direct-attachment of storage devices (like drives and SSDs), eliminating the SDS software latency and cost. CPI also doesn’t require a hypervisor to run an SDS layer or workloads. Instead, it creates bare metal server instances that can support containers or a hypervisor if desired, reducing software licensing and hypervisor lock-in.
Composable Infrastructures are stateless architectures, meaning they’re assembled at run time, and can be controlled by 3rd party development platforms and management tools through APIs. This improves agility and makes CPI is well suited for automation. For more information see the Technology Insight paper “Composable – the Next Step in Integrated Infrastructures”.
The amount and diversity of technology available in infrastructure products can be overwhelming for those trying to evaluate appropriate solutions. In this blog we discuss pertinent topics to help IT professionals think outside the checkbox of features and functionality.