Data Centers and 5G: How the new generation of mobile networks is changing infrastructure demands

With the development of mobile network technologies, we observe how each new generation changes our habits and communication requirements. While previously 4G networks mainly helped us stay connected and access the Internet, 5G opens completely new horizons.

This generation of networks not only improves communications but also significantly enhances interaction between devices, as well as Internet of Things services and artificial intelligence. In this regard, data centers (DCs) must adapt to new conditions.

Data center service providers need to think ahead about how to update their infrastructure and what new approaches to use for data center management. This includes both large data centers and small edge micro-DCs. Additionally, it’s important to explore automation possibilities to ensure the smooth operation of 5G networks. In this article, we will examine how exactly 5G will change the requirements for data centers and what needs to be done to successfully adapt to these changes.

Three basic services of 5G networks can be identified, upon which all the diversity of their services is built:

• eMBB: Enhanced Mobile Broadband. While in previous generations the main services were voice transmission and Internet access, in 5G the main services are high-resolution video, up to “immersive video,” as well as processing large volumes of data. The volume of video data is orders of magnitude higher than the volumes of data transmitted in 4G networks.
• mMTC: massive Machine Type Communication. 5G networks can provide not only ultra-high-speed connections but also, through special protocols, relatively low-speed numerous Internet of Things devices.
• uRLLC: ultra-Reliable Low Latency Communications. For critical 5G services provided with high reliability, communication with ultra-low latency will be required. Instead of the usual 10-50 milliseconds, 5G networks must be capable of providing information transmission with a delay of no more than 1 millisecond.

These three main groups of basic 5G services should provide an enormous variety of services in 5G networks: autonomous vehicles; telemedicine; critical applications requiring low latency and high reliability; smart homes and smart cities; high-resolution immersive volumetric video; processing huge amounts of information from Internet of Things sensors and detectors, and much, much more.

To support 4G networks, data center operators simply added more equipment: increased the number of servers, routers and switches, and engineering systems. This led to increased power consumption and cooling costs. Scaling such an architecture is very costly and time-consuming. In Internet of Things (IoT) networks, there are many devices requiring constant connection and generating large volumes of data. This makes it difficult to manage and scale 4G services for telecom operators and corporate networks, resulting in significant traffic growth that requires substantial server resources for processing.

Therefore, data centers began using virtualization tools (virtual machines and containers) to reduce the number of physical servers, increase their specific workload, and optimize the system infrastructure.

Requirements for data centers for 5G are changing under the influence of a significantly expanded range of services based on the basic services mentioned above. However, 5G networks also involve many technological changes that enable the provision of an expanded range of services. Let’s highlight the main ones:

1. NFV/SDN (Network Function Virtualization/Software Defined Network): Virtualization of network functions and software-defined networks. Traditional networks of previous generations 2/3/4G were designed and built based on specialized hardware devices. The deployment of such “monolithic” network elements led to long design and commissioning cycles, slowing down the market introduction of new products and services.

In 5G networks, instead of physical equipment, virtual network functions (VNF) are used, which emulate the operation of physical equipment based on standard and relatively inexpensive server equipment in data centers (COTS – Commercial Off The Shelf). Therefore, it can be said that 5G network operators not only “use” external data centers to host various systems (for example, OSS/BSS), but they themselves “live” in data centers, using virtual network functions (VNF) instead of physical equipment (see figure on the right).

Network virtualization based on NFV organically entails the following two changes, the technological basis of which are also data centers.

2. Network Slicing: Virtualization allows not only flexible redistribution of data center computing resources in the operator’s core network but also the allocation of logically isolated virtual layers (slices) of the network, so that resource load in one slice does not affect service quality in another.
3. Edge Computing: In 5G networks, data centers are used not only at the network management and application level and for building the operator’s core network and aggregation network but also at the access level (modular and containerized data centers).

Thus, innovations in data centers for 5G are closely interconnected and can create advantages through the synergy of these innovations.

1. AT&T

   AT&T is the largest telecommunications operator in the US and one of the largest fixed and mobile operators in the world. Back in 2015, due to the explosive growth of traffic caused by the proliferation of smartphones and tablets and the upcoming transition to 5G networks, a “complete rethinking of the network concept” was conceived. This rethinking based on virtual network functions enables adequate network capacity growth as traffic increases, while network development costs remain at an acceptable level. To date, AT&T has managed to transfer most of its network elements to NFV technology, which operate in data centers as software entities. Moreover, not only AT&T’s telecom functions are virtualized in data centers, including cloud-based ones, but also IT functions (billing, management, databases, etc.), which were previously maintained by the operator as a separate corporate IP network. Therefore, AT&T no longer has separate departments for IT and commercial telecom networks, which has allowed for even greater reduction in internal costs.

2. China Mobile

   China Mobile, the largest mobile operator in China, recognized the capabilities of NFV technology for rapid scalability and implementation of new and improved services about 10 years ago. According to the company, “China Mobile has built a colossal NFV network with 178 thousand servers supporting more than 370 million 5G users and 270 million IMS (IP Multimedia Subsystem) users.” China Mobile built a pilot NFV network in 2015 and put it into commercial operation in 2018. To date, China Mobile has virtualized more than 85% of all network functions. As a result, telecommunications-level service reliability of 99.999% has been achieved.

3. DOCOMO

   The largest Japanese operator DOCOMO has been running a project since 2015 to transition its network to virtual functions. Currently, more than 90% of the operator’s network has been virtualized, utilizing over 400,000 virtual processors (vCPU) (see figure below)

Thus, the transition of operators’ core networks to NFV technologies when building 5G networks is an inevitable trend. Using data centers for virtualizing network functions of operators is a more advantageous solution due to factors such as:

• Reduction of capital expenditures for the telecom operator, because instead of purchasing equipment for hardware network functions in traditional equipment, the operator can use virtual network function (VNF) templates, compose required network services and applications from them, and run them on standard servers in data centers in the quantity required for uninterrupted provision of services and applications to users (according to DOCOMO, in their NFV project, CAPEX reduction was more than 10%);

• Reduction of operational costs, as it eliminates the need to hire highly specialized and expensive specialists to maintain hardware telecom equipment, its upgrades, and maintain operability; in addition, operational costs for maintaining virtual infrastructure based on standard data center equipment (COTS) are usually much lower than for maintaining specialized equipment;
• Expansion of capabilities in developing and launching new services and applications, as it eliminates the need for a long cycle of designing, purchasing, and launching equipment for new functions based on specialized equipment, while for virtual functions this process is significantly accelerated and simplified, since newly developed virtual functions require only programming and debugging on standard server equipment in data centers.

These are just the main advantages of transitioning operator networks to NFV, not counting a number of indirect and consequential advantages.

Considering the urgent need for telecom operators to transition to NFV technology in data centers, a question arises – how to implement such a transition? Should the operator build its own data centers to create the necessary NFV infrastructure, or use the services of third-party data center providers through the “colocation” model?

The choice for operators, at first glance, seems obvious – they need to build their own data centers. This is a popular approach, often referred to as “walled garden” in technical literature. Typically, business leaders subconsciously strive to have and use “everything of their own.”

However, whether such an approach is optimal remains a significant question. The costs of hardware solutions for communication networks, which decrease with NFV implementation, are replaced by costs (and not small ones) for building their own large data centers (carrier-specific DC).

From an economic perspective, a more optimal approach appears to be using third-party (neutral) data center service and infrastructure providers (carrier-neutral DC) that offer equipment hosting services. There are several reasons to choose a neutral data center, such as:

• Even greater reduction in capital expenditures: costs for data center construction, server equipment acquisition, and its technical support in the case of outsourcing to a carrier-neutral data center are replaced by operational costs for renting its virtual infrastructure.
• Flexibility, which allows operators to choose data center service providers based on factors such as price, service quality, and reliability;
• Performance – neutral data centers typically have multiple connection points for different operators, with specialized routing equipment to ensure low data transmission latency.
• Resilience – operators using neutral data centers receive higher service availability, approaching 99.999%, as they are not dependent on a single operator.

There are other advantages of carrier-neutral data centers as well.

Neutral data centers offer many advantages for large-scale use, but at the moment, the first data centers for public 5G networks will likely be the telecom operators’ own data centers. This is the path followed by the aforementioned operators AT&T, China Mobile, and DOCOMO.

The strategy for neutral data centers will probably focus on projects implementing private 5G networks. Many large enterprises are interested in departmental 5G networks because they see them as the best combination of both network types: high performance and low latency (public 5G) combined with reliability and flexibility in service provision (private 5G). According to ABI Research, the market for private cellular networks will grow to $100 billion by 2030, compared to $7 billion in 2023.

For example, Equinix, one of the world’s largest data center service providers, offers enterprises building private 5G networks the option to place their 5G User Plane Function (UPF) in Equinix IBX® data centers using the colocation model. Since the UPF function is the foundation for aggregating and distributing 5G traffic, it’s crucial for enterprises to position their 5G UPF at the edge of the data center network to ensure low-latency connectivity and operational efficiency for their corporate network. Equinix data centers are available in more than 70 cities worldwide, giving their clients great flexibility in choosing connection points.

The infrastructure of data centers (servers, storage systems, local and global data networks, management systems) faces many new requirements in the context of 5G network deployment. Among the main requirements are the following:

1. Support for virtualization technologies (NFV) and software-defined networks (SDN/SD-WAN) to enable the replacement of hardware communication network equipment with virtual network functions (VNF), thereby increasing the efficiency of data center computing resources, reducing energy consumption, and enhancing management automation.
2. Support for distributed data processing infrastructure to create Mobile Edge Computing (MEC) architecture, which enables flexible redistribution of radio access network resources depending on user traffic load in individual cells and reduces traffic in the aggregation network and 5G Core.
3. Support for artificial intelligence (AI) and machine learning technologies, which will allow for operational and automated management of distributed computing infrastructure of data centers, core level, aggregation, and edge network, to ensure efficient resource management and provide new services and applications to the entire spectrum of 5G users.
4. Availability of a flexible and automated NFV service and resource management and orchestration system: MANO (Management and Orchestration), which is necessary for the rapid formation of services on demand and without delays, as well as for efficient distribution of workload on computing and network resources of the distributed data center network in 5G.
5. Ability to connect and manage modular and container data centers at the network edge (Mobile Edge) and flexibly manage their resources in a distributed network, which allows combining the computing infrastructure resources of distributed data centers into a single managed and automated network.
6. Availability of service and application development platforms, as well as the ability to provide them to users through APIs, which enables 5G network operators to quickly develop and provide new services and applications according to market requirements, thereby increasing the competitiveness of the 5G operator.
7. Ability to interact with third-party development platforms and Over The Top (OTT) service providers for conducting joint and mutually beneficial business, thus expanding the range of services and applications provided, even for those user groups for which the operator’s own service development may be impractical due to small market size or specific requirements of niche markets or specialized applications.

Neutral data centers represent one of the foundations for building and operating 5G networks, demonstrating true convergence of information and communication technologies (ICT). In the long term, telecom operators will increasingly consider these facilities and use Colocation services rather than invest in building their own data centers.

For large enterprises with private communication networks (private 5G), collaboration with neutral data centers, such as IXcellerate, will also significantly increase resource efficiency, improve performance and responsiveness in providing new services. Thus, neutral data centers are becoming a key element of a successful strategy in the 5G era.