There is another level in virtualized environments that surpasses the expectations of all other infrastructures, and these are hyper-converged systems.
A converged system is a set of components (servers, storage devices, network) combined in an infrastructure solution powered by software and a unified management platform. Among the solutions that permit this unified management, the most recognized are Cisco UCS and IBM Flex. Yet even though the range of services is controlled from a single management console, the hardware infrastructure components remain separate.
A hyper-converged system includes the same components (server, storage and network) in a single chassis with a layer of management and automation. In hyper-converged systems, the design of the environment is devised in a cluster. In this environment, each node has access to the storage of all the other nodes.
This solution allows access to data and server applications through a single interface.
Since all the components are in the same housing, a better density is obtained with a hyper-converged infrastructure. The increased density often allows more operations in less physical space.
For example, in a converged infrastructure constructed in an enterprise 42U rack, the storage could use 6U and the systems 8U, for a total of 14U. Conversely, in a hyper-converged system, these components would fit in a space of 4U.
This space saving equally reflects a decrease in the release of heat from the devices and therefore a diminished need in terms of cooling, hence electricity.
What contributes to the high performance of hyper-converged infrastructures, among other things, is the distribution of VMs in a FileSystem managed and optimized by the cluster. For example, Nutanix is designed this way and allows for the integration and direct relationship between compute and storage. This allows quick access to data due to their proximity to compute. Consequently, the movement of data between the various tiers of storage is also optimized.
Of course, the speed of data transfer depends partly on the type of disc used. The ultra-fast SSD or Flash type drives will provide the best performance, while the mechanical drives (spinning drives) will be slower. However, these two technologies are part of a hyper-converged system.
The very design of a hyper-converged infrastructure is focused on simplicity. Given that compute, network and storage elements are all integrated in one housing, this avoids having a person who will manage each component separately. The distributed FileSystem alone manages all exchanges between the different tiers of storage.
Compared to a traditional or converged infrastructure, a hyper-converged environment will bring several types of gains:
• Few components to support
• Simplified deployment
• Predictability of performance
• Easy growth with the installation of additional nodes
• Automated configuration of new nodes
• Ease of scalability based on business needs
With most traditional integrated systems, these actions are multiplied. If you add a component, it must be configured separately and then connected to the other components.
There are many myths and misconceptions around hyper-convergence since it is a type of environment that has only been developed in recent years. These misconceptions revolve mainly around the cost of implementation, the complexity and who these solutions are for.
The choice of a hyper-converged system relies as much on strategic considerations for optimisation as it does on technical features.
© freshidea - Fotolia.com