What is Converged Infrastructure?
The definition of Converged Infrastructure (CI) can be categorised into different ‘flavours’. In summary, the term defines an approach where storage and compute are tightly coupled, and managed within a software-defined paradigm that offers policy driven provision of resources. Different vendors provide their own ‘flavour’ of CI, ranging from an infrastructure package that is optimised to an application, such as Oracle’s Exadata system, where a single rack contains storage and compute modules that are pre-integrated, cabled and pre-configured hardware that is tuned to deliver optimum results for a specific application. The administrator has only to power on the system and connect it to their current environment.
Early Converged Infrastructure solutions from vendors such as HP and VCE (a consortium of VMware, Cisco and EMC) took a similar approach, with a rack of modular pre-integrated and configured systems, optimised for general purpose workloads. The approach of these packaged CI systems also allows for automated expansion, where a new unit can be rapidly integrated with the existing CI system. However, packages from different solution providers are often not directly compatible with each other, as they may often depend upon bespoke management or control software that abstracts the individual underlying modular components from the administrators.
It is a popular practice to create a converged infrastructure in software alone, where modular components are integrated with software that is abstracted from the individual complexities of components. A primary component of this is to utilise a virtual SAN, where disk storage directly attached (or inside) the compute infrastructure is presented as capacity to be utilised by virtual machines, and software defined storage policies control the placement of data and virtual machines dynamically. The integration of subsequent infrastructure modules in this CI approach is then driven by software policies that streamline expansion, such as vSphere Host Profiles and AutoDeploy.
A growing and future ‘flavour’ to Converged Infrastructure has been granted the term “HyperConverged”. The differentiation of this approach to other CI flavours is that smaller modules include pre-built virtual SAN and virtual networking, with management and control that provides turn-key resources. It can be considered to be “a datacentre in a box”, with automated self-discovery of nearby modules for rapid expansion. The drawback from this new paradigm could be from their biggest benefit – the pre-integrated modules are a closed system, requiring the usage of the included packaged management and control systems, and the package will include all the software licensing and integration, instead of providing a logical upgrade path from existing systems.
VMware’s EVO:Rail reference architecture and software overlay is a HyperConverged package with standardised modular components that are available from multiple vendors, closely integrated by software that provides a wizard for initial deployment. The first EVO:Rail deployment can be up and running in less than an hour from unboxing to being available to run workloads, making this a great choice for branch offices, new companies and startups, or simply for rapid and dense deployment. Subsequent modules will detect each other, making expansion and new deployment very rapid.
