In the past, data centre computing was characterised by large, integrated resources called mainframes. As shared computing resources became more decentralized and affordable, these resources split into distributed computing layers of servers, storage, networks and software. To curb the proliferation of distributed IT sprawl, these computing layers were pooled and virtualized for maximum efficiency and minimum costs. Today, to be able to rapidly deploy new pooled resources, the computing layers are delivered in a single chassis either from a single vendor or as reference architectures. This is what is now commonly known as converged infrastructure.
The figure below outlines the four types of system architectures introduced above and their key architectural differences and selling points.
Integrated – Orchestration, computing, storage and networking are tightly integrated in a single box. System upgrade and growth is through replacement of the entire system. The architecture is designed for a specific purpose or workload.
Distributed – Computing and storage are in separate system blocks orchestrated separately and connected through networks. System upgrade is through replacing component blocks. System growth is through adding blocks. The architecture is designed to enable growth and scale out of multiple workloads.
Pooled – Computing, storage and networks are in separate resource pools consisting of blocks and orchestrated separately. System upgrade is through replacing blocks within a pool. System growth through adding blocks to a pool. The architecture is designed to enable efficient scaling and growth of multiple workloads.
Converged – Computing, storage and networks are in separate resource pools jointly orchestrated and in a single chassis. System upgrade is through swapping components within the chassis. System growth is through adding components to empty chassis slots. The architecture is designed to support rapid deployment of multiple workloads. Within converged infrastructures the pendulum swings back to the integrated system architecture with vendors also offering fit for purpose versions for specific workloads. Case in point is IBM PureSystems that offers PureFlex as the generic flavour and PureApplication for cloud and pattern workloads and PureData for big data analytics.
Who is selling converged infrastructure?
Continue reading on the website of The METISfiles