Hyperconvergence and Infrastructure Composability

1. Summary

There has been an increasing demand for simpler ways to provision and manage infrastructure resources over the last decade. Traditionally, enterprise IT infrastructures were designed around Network-attached storage (NAS) and Storage Area Networks (SAN) on dedicated FC equipment, with network connectivity provided by separate Ethernet switches and large scale-up compute systems. Hypervisors, private cloud, and other technologies have changed the way the compute layer is organized and lately the datacenter design approach has been impacted as well.

Converged infrastructure (CI) was the first attempt to simplify traditional models. Pre-configured building blocks, based on converged Ethernet, traditional shared storage, and a number of x86 servers were integrated together and shipped as a whole with specific management tools aimed at simplifying provisioning and management. Procurement and provisioning is simplified, but at the cost of granularity and resource utilization efficiency.

Hyper convergence (HCI) has further improved this concept and efficiency. With the help of software-defined solutions for storage, and later networking, relatively small bricks with onboard compute and storage can be connected together to form a large cluster usually governed by a single hypervisor and additional software tools. It offers a highly simplified infrastructure model which provides several benefits to enterprises of all sizes.

Composable infrastructures take granularity and efficiency to the next level. By building pools of resources that can be distributed and reconfigured where necessary almost instantly, it is now possible to attach the right storage resource, both in size and performance, to the compute node as needed – improving efficiency and overall system provisioning while keeping costs as low as possible in large scale infrastructures. Composability is focused on bare-metal resources, without needing specific software layers such as hypervisors.

In this report, we analyze several aspects of hyperconverged and composable infrastructures including:

• The benefits introduced by Hyperconvergence and infrastructure composability
• Applications and workloads that can take full advantage of a composable infrastructure
• Comparisons between HCI and composable infrastructures
• The role of automation
• Enabling storage protocols
• Hardware- and software-defined approaches to hyperconvergence and composability

Key findings:

• HCI is now widely adopted by enterprises of all sizes and the game is quickly shifting from on premises installation to hybrid cloud and edge integration.
• Composable infrastructures are quickly gaining interest because of their ability to speed up the provisioning process of bare-metal resources, which is fundamental for both a growing number of applications and workloads in many different types of organizations
• Composability offers unmatched granularity and efficiency, as well as performance, enabling organizations to run big data and HPC clusters on their premises for the time necessary to get single jobs done and repurpose the cluster quickly for other tasks.
• New technology, such as NVMe for example, is helping to shifting from traditional storage paradigms to innovative architecture designs, more suited to swiftly meet demanding business requirements which need Big Data Analytics and other CPU and data intensive applications.