Dynamization of data centers - How can this be implemented in cabling?

Data centers have increasingly become the economic and social hub of many activities. Business processes, online shopping, gaming, or traditional e-mail correspondence could not work without them.

The corona pandemic has placed particular demands on digital centers. Within just a few weeks, data centers have become the platform for school and university operations, among other things. In order to reduce the incidence of infection even further, the healthcare system, police and disaster control agencies are also turning more and more towards digital exchange. And last but not least, companies have been and still are requiring the services of data centers, as more and more employees were and are working in the home office due to the lockdown and the potential danger of infection.

Virtualization and software defined data centers enabled data centers to adapt quickly and flexibly to the increased load and provide the necessary capacities. However, dynamic adaptation can only be successful if the underlying infrastructure can keep up. The infrastructure must be performant, flexible, fail-safe and expandable so that the high-performance software solutions, network operating systems and management solutions can unfold their full potential.

Classical typology as a starting point for modern developments

The conventional IT infrastructure of a data center is based on numerous components performing different tasks. They are not directly involved in data processing and the processes required for it - but without them, nothing in this environment will function at all.

Basic components are responsible for the air conditioning in the data center, but also provide the power supply and distribution. Additional solutions are required for the various security tasks. The spectrum here ranges from access control to firewalls. Other areas that are part of the data center equipment include building management and various services, as well as telecommunications cabling. The interaction of the various units must work together flawlessly in order to provide sufficient capacity to ensure smooth data center operation without bottlenecks.

As requirements grow, so does the desire for a dynamic data center. In contrast, however, the static infrastructure in many areas offers little room for flexibility. Data center expansion itself is not enough, because in many cases there is simply not enough space. Therefore, modularity is required for the components. When it comes to power supply, modular uninterruptible power supplies (UPS) and modular, manageable power distribution systems with sufficient redundancy have proven their worth for years. Even the built-in cooling systems with intelligent control usually have sufficient reserves to cushion fluctuations. But how does the cabling fit into the principle of dynamization?

Spine-leaf architecture outshines classic structures

Cabling in data centers has been based on a hierarchical architecture for decades. It largely follows the requirements of the DIN standards EN 50173-5 and EN 50600-2-4, but also has disadvantages resulting from the hierarchical structures. These include long transmission paths, for example. In some cases, this requires a large number of switches between the servers, which inevitably leads to latencies and latency differences. In extreme cases, the transmission rates deteriorate, which puts pressure on performance. In analogy to a further expansion of the existing server equipment, additional cabling is needed. "Bottlenecks" in the data center infrastructure are the consequence. The failure of central core switches poses a further threat to performance even if they are designed redundantly.

Since applications are running on virtual machines and data traffic continues to increase, the issue of "bottlenecks" is increasingly becoming the focus of attention. In order to ensure smooth processes, blockages within the network must be eliminated. The so-called spine-leaf architecture offers a remedy. The underlying approach: Every leaf switch is connected to every spin switch. This creates a tightly meshed network.

Conventional realization of the Spine-Leaf connection scheme — Conventional spine-leaf architecture

The data stream in this meshed concept is randomly selected and evenly distributed. In case of imminent overload, an alternative path is automatically selected. The data traffic between leaf switches operates over one spine switch at maximum and therefore uses existing latencies. However, there are also disadvantages with this structure. For example, unwanted spanning tree procedures cause dropouts that can only be prevented by routing protocols and their careful configuration in the switches. In addition, the operative effort is massive, which is caused by the high number of physical connections and the complex cross-connection topologies. This in turn has a negative impact on scalability.

Dynamic cabling with the CrossCon^® solution

In order to optimize the spine-leaf architecture in the data center and to make it more dynamic in terms of scalability, Rosenberger OSI and FiberCon have combined their expertise in the field of fiber optics and interconnection technology in a joint project and developed an MTP^®/MPO version of the CrossCon^® system from FiberCon.

FiberCon's patented CrossCon^® system ensures standard-compliant, structured, yet flexible data center cabling. Its novel plugging scheme allows each connected rack terminal to communicate with any rack terminal of the entire cross-connection scheme in the data center. In contrast to the classic design of a spine-leaf architecture, there is no need for complex cabling here, since the signals are crossed within the CrossCons^® and are only routed to and from the CrossCon^® with patch or trunk cables. Thanks to this innovative signal routing, the documentation of the cable routing can be significantly improved and the number of required plugging processes can be reduced. Complex work processes during initial installation and subsequent expansion of additional routers are bypassed, and statistical error sources are reduced. Thus, costs are reduced and administrators are relieved of time-consuming routine tasks. This is why a high degree of scalability can be achieved with the CrossCon^® connection core.

Spine-Leaf connection scheme with CrossCon® system — Spine-leaf cabling with CrossCon^®

The advantage of the MTP^®/MPO version of the CrossCon^® system lies in its independence of manufacturers. This means that new configurations and extensions can be easily implemented in any data center environment. In addition, it offers considerable space savings: MTP^®/MPO connectors can accommodate up to 72 fibers, which means that the space on the PCB and in the rack can be better utilized. With the development of this solution, a further step towards the dynamization of data centers has been taken. This makes them ready for new challenges.

Are you planning to make your data center topology more flexible or have you already gained experience with the spine-leaf architecture?

Condividi articolo

Articoli simili