22.08.2020 I by Henrik Hasenkamp
Hyper-converged infrastructures are intended to combine the qualities of cloud solutions with the advantages of a dedicated data center. What is behind this technology and when is its use appropriate?
A company's IT infrastructure is traditionally made up of several components: servers process data, storage stores it, users use applications on their end devices and desktops, and the network in turn connects everything together. Even if the individual components are necessarily connected with each other and only together result in a functioning IT, they are managed and configured separately. Administration and configuration is thus comparatively complex and costly, and changing IT requirements usually require specific adjustments to several components, thus increasing the frequency of errors. Moreover, modern and automated IT concepts cannot be mapped at all, since the individual components cannot be controlled via modern programming interfaces.
In practice, the individual components are therefore growing together more and more. They are configured and combined with each other in such a way that they interlock closely - so-called convergent infrastructures are created. Often several components are combined to form an appliance, i.e. a combined system of computer hardware and software specially optimized for this hardware. This centralizes parts of the administration and management, but at the same time reduces the usability of such an appliance to mostly pre-defined scenarios. The advantages of administration and management are thus followed by the disadvantage of less flexibility.
Hyperconvergence through virtualization
Despite this close coordination, however, the individual systems in convergent environments remain just that: individual systems consisting of a hardware level including the respective software. Thus, a convergent appliance often combines several proprietary approaches. Here, hyper-convergence goes one step further: the device intelligence that used to sit in the individual components is replaced by an intelligence that spans the entire IT infrastructure. In concrete terms, this means that the entire IT infrastructure will be combined into one large overall appliance. A comprehensive management software takes over the central control and administration of all components.
The basic technology for Hyper-Converged Infrastructures (HCI) is virtualization, which is already actively used in most data centers. This means that not every server really needs a physical counterpart, but rather several virtual environments run smoothly on the same hardware. A hypervisor forms the abstract intermediate level and dynamically distributes the provided resources as needed. The available capacities can thus be used much more effectively and flexibly. Each request is allocated free capacities accordingly. In principle, it does not matter on which hardware system the resources are actually physically located.
In a modern HCI architecture, the hardware provides all functionalities - from server to memory to network - fully virtualized. The hypervisor acts as an interface and management software. It organizes the virtualization of all available resources and the load balancing of requests. Storage and network become software-defined components. This means, for example, that no external storage system needs to be connected, but that all distributed, individual storage units are combined into a virtual overall system. Similar to a cloud, the user is largely not interested in which hardware resources are used where at any given time - what he needs is flexibly compiled as required from the entire resource pool.
Advantages and limitations of HCI infrastructures
The decoupling of hardware and software brings several advantages. HCI systems are usually offered as a complete package, pre-configured by the manufacturer - in a sense as a plug-and-play data center. In addition, HCI infrastructures are much easier to scale than classically structured IT environments and avoid the need to maintain oversized server hardware. Additional systems and components can also be integrated particularly easily. The hypervisor includes the resources in its pool and makes them immediately available to users. In this way, new capacities can be implemented quite quickly while avoiding isolated silo solutions.
Virtualization also has a positive effect on the performance of all systems, because all resources can be allocated dynamically and, if necessary, also ad hoc and on a correspondingly large scale. The same applies to availability: The failure of individual parts can be compensated for almost arbitrarily from the resource pool. What's more, the comprehensive structure makes it much easier to automate cross-departmental processes and map them uniformly on the IT infrastructure. The consistent structure and central management are particularly helpful here. In addition, the HCI infrastructure can also be managed by a specialized service provider if required. Precisely because the IT comes in the form of an HCI as a complete package, it can also be easily obtained as a managed service or finished infrastructure service from a cloud provider.
But HCI systems do not prove to be suitable for every company and every application in practice. While the complexity of the infrastructure as such is reduced, that of the management software increases. Because now all systems are virtualized and managed with a central solution - this requires extensive and comprehensive know-how.
For various reasons, the change from an existing IT paradigm to HCI technology is not easy. HCI is not suitable as a supplement for rare, temporary resource bottlenecks - it is too oversized for that. In addition, while it may be an advantage to purchase an almost turnkey system, it is also a disadvantage: Because ready-made complete packages rarely actually cover all individual needs. In addition, one often becomes technologically dependent on a single provider.
Costs are also a sensitive issue. On the one hand, the ongoing administration and operating costs are reduced considerably due to the usually lower administrative effort and better scalability. On the other hand, this is offset by high investment costs - especially if the HCI infrastructure is to be purchased rather than rented. Therefore, companies should definitely check in advance whether an expansion of the infrastructure may be sufficient or whether cloud options are also possible as cheaper and more flexible alternatives.
Is it worth it, is it not worth it?
Whether the advantages outweigh the disadvantages must be decided by the specific application in the company. For example, an HCI system can be of interest if a data center is needed for a foreseeable workload that can be deployed quickly. Moreover, convergent and hyper-converged infrastructures are not mutually exclusive. For example, an HCI can also be tested beforehand on a dedicated project in the company. The experience gained from this can then be used to plan how the IT landscape should develop in the future. It is therefore difficult to make a blanket statement, but it is generally true that the economic advantages of an HCI solution bring exponential cost benefits as the size of the company's own IT infrastructure increases.
All in all, HCI technology can help many companies to modernize their IT, to better control it and to significantly accelerate many processes. Thanks to virtualization and cloud-like services, the infrastructure gains both flexibility and dynamism at the same time and thus a significantly increased ability to react to changing business requirements.
The original article in german can be found here.