Engineers from La Salle-URL share the latest news and projects in the field of network solutions in telematic engineering.

23 February 2021 | Posted by userDataCenter

Hyperconvergence

The evolution of data processing centres (datacentre) over time has been aimed at optimizing hardware resources in order to save energy costs, as much as material and multiple devices to buy, just like the space that the same occupy. Following this paradigm, it is right to determine that the evolution is to achieve more and more centralized devices and products that will allow simplicity in the field of infrastructures. Due to this approach, concepts have emerged to the convergence, or its evolution, hyperconvergence (HCI). 

 

THE ORIGIN  

Initially, datacentres used non-convergent infrastructure models. These were characterized by segmenting the storage resources in their own network (SAN), isolated from the other network functionalities granted by the network cards (NIC). This fact led to the extensive deployment of devices for the sole function of having a server and storage.  

Following this example, having the topology previously discussed, it would be necessary a physical server, a storage device, as well as the switch used for NICs, and the internal switch of the SAN, which makes a total of at least 4 physical devices. 

 Scheme of a non-convergent infrastructure.

Diagram 1: Scheme of a non-convergent infrastructure. 

 

WHAT IS CONVERGENCE? 

As a result of this drawback, as well as the cost and excess of devices within the infrastructure, this architecture evolved into a convergent model. This one, improved with respect to the previous one when uniting and offering unique connectivity that combined the generic network traffic with the traffic storage of the SAN network. Following this purpose, an optimization was obtained in the purchase of devices by for example being able to unify both traffic. 

Following the previous example, a convergent infrastructure would only require a single switch instead of 2, requiring a minimum of 3 devices, which leads to an optimization of 25% over the previous example.

Scheme of a convergent infrastructure

Diagram 2: Scheme of a convergent infrastructure. 

 

WHAT IS HYPERCONVERGENCE? 

the evolution of data centres tends to centralize and simplify their infrastructures, thus making the most of resources per device, as well as making them multifunctional. This is the purpose of hyperconvergence, since for example, if the servers already have built-in storage, why would an additional external device be used if the same can be used making a distributed file system (DFS).

A sum of resources is obtained by implementing a distributed system of equipment that combines its resources creating a logical entity with the total of these. Thus, by modifying the total equipment, the only impact would be the addition or subtraction of resources from the total sum. 

Following the above example, in this case, we could give up storage devices, as they would be integrated with the same servers, so only a minimum of two devices would be needed, the server and the switch (following the previous case) that would intercommunicate them, which leads to an optimization of 33% with respect to classical convergence, and optimization of 50% with respect to non-convergence. Remember that these percentages are taken from the example discussed, and do not refer to the average percentage of any study. 

Scheme of a hyperconvergent infrastructure

Diagram 3: Scheme of a hyperconvergent infrastructure. 

 

PROS

As mentioned above, hyperconvergence is a very promising alternative to consider in terms of convergent and non-convergent technologies, mainly because a three-layer architecture is expensive to implement, in addition to scalability difficulties and complexity of operations.  

However, there are numerous benefits to its use, such as:  

  • Maximized use and optimization of equipment 
  • Automation of tasks more simple and agile, which allows to implement this architecture in a few hours, and once deployed allows saving on processes and simplifying the management of tasks 
  • Great scalability and flexibility, being able to adapt to the needs of the companies with the use of an integrated layer 
  • Solving problems of compatibility between security systems, storage, management of servers and networks, since now they’re found integrated 
  • Reduction and efficiency of costs maintenance 
  • Possibility to increase the speed in certain aspects, especially in duplication processes 

 

POINTS TO IMPROVE 

However, a transition to a hyperconvergent architecture introduces new challenges and certain disadvantages over other technologies, in this case, its predecessors, with the first economic challenge being its initial investment. In addition, the use of a hyperconvergent system means opting for an integrated solution that depends solely on a supplier or manufacturer, a dependency that can cause devastating effects in the event that the contracted operator has operational problems or is not reliable enough, a problem known as vendor lock.

On the other hand, another disadvantage of this architecture is the cost in the performance that generates control of distributed resources previously installed that are not thought specifically for an infrastructure of hyperconvergent systems, since it bets by interoperability between its components, sacrificing the performance of individual devices.  

 

CURRENT MARKET POSITIONING 

Currently, one of the main tools used by consumers, manufacturers, integrators, consultants and other sectors that hire services and technological solutions is what is known as the magic quadrant of Gartner, a tool that lets you know at what point of innovation and level of development are the tech companies located in the world market. This graph is mainly based on two criteria (represented on its two axes): the completeness of vision and the ability to execute, generating 4 different areas to where locate the companies involved in the evaluated technology, which are: the challengers (they have good execution of the business and are powerful but do not have yet demonstrated an understanding of market needs); the leaders (they are well-positioned for the future and their business vision is aligned with the current market); the Niche players (not characterized to surpass their competitors but behaving successfully on a specific market segment); and finally the visionaries (they understand the needs of the market but are not yet able to implement it). Thus, the companies on the first quadrant are the ones that provide the best services of the technology in question. 

By using this tool to analyze the market in hyperconvergent infrastructure technologies, it can be seen that the most successful are Nutanix and VMWare, which repeat their privileged position in the leader's quadrant, while most other providers struggle in the market niches. The solutions of these two companies provide complete integrity of processing, storage, network and memory in a single container or cluster with integration in the cloud and also providing a backup and replication system. 

Representation of the Gartner magic quadrant for hyperconvergent infrastructure systems with data from December 2020

Diagram 4: Representation of the Gartner magic quadrant for hyperconvergent infrastructure systems with data from December 2020

 

Jaume Campeny i Alba Massa
Share

Add new comment

CAPTCHA
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.
4 + 5 =
Solve this simple math problem and enter the result. E.g. for 1+3, enter 4.