5G Means Data Center Platforms Must Evolve
Image Source: buffaloboy/Shutterstock.com
5G is the next frontier of cellular connectivity, but the new generation promises more than just faster download speeds and lower latency. The higher bandwidth and more comprehensive coverage 5G is expected to deliver will open up various new use cases for connectivity beyond just cellular phones, ranging from laptops and handheld Internet of Things (IoT) devices to automobiles and large-scale industrial implementations. Industry forecasts predict that 5G will gain over 1 billion subscribers by the mid-2020s as the migration from 4G devices that aren’t compatible with 5G networks takes place. However, transitioning to 5G will require a major investment in new cellular infrastructure. 5G architecture will change substantially from preceding implementations. Driving these changes are some of the key characteristics of 5G that have evolved from the fourth generation. We’ll review these key characteristics and then see how these will affect the data platform architecture for a typical 5G system, and explore the implementation options for various levels of data platforms and identify typical implementation options. We’ll also look at an example mid-range data platform in detail to identify key design choices and trade-offs.
5G Characteristics Drive Implementation Architecture
Low-band 5G cell towers operate on a frequency range similar to 4G cellphones (from 600MHz to 850MHz and provide similar range and download speeds (30Mbit/s to 250Mbit/s). As a result, low-band 5G is already being phased out in many areas of the world. Mid-band 5G towers employ microwaves of between approximately 2.5GHz and 3.7GHz, significantly increasing download speeds to the 100Mbit/s-900Mbit/s range while expanding coverage range by several miles. Mid-band 5G towers are already the norm in big cities and other high-population areas and could soon become the worldwide standard.
High-band 5G currently operates in the 25GHz-39GHz range and delivers download speeds similar to cable internet service, which is about 1Gbps. High-band 5G does have limitations, however. Operating at the 25GHz-39GHz range is on the low end of the millimeter-wave (mmW) band. mmW has a more limited range than microwaves, meaning high-band 5G will require a greater number of smaller cells to cover the same area as mid-band 5G. Physical obstructions such as walls or home appliances can also limit high-band 5G connectivity. mmWs also don’t negotiate solid objects well. High-band 5G is also much more costly than lower-frequency technology. As a result, high-band 5G might be limited to large, relatively open facilities such as concert venues and sports arenas in the near future.
The Pyramid of 5G Data Platforms
Factors such as coverage range, required download speed and cost efficiency have to be considered when determining the 5G hierarchy level for a specific implementation. The 5G distributed data platform places data processing, storage, and communications at various levels of the architecture hierarchy to optimize cost, power, network performance, operating distance, and user features. Closest to the network’s edge are small pico-platforms that cover small distances (tens of meters) within a building or facility. Common examples include building automation, security, factory floor module monitoring, and control. On the level above edge devices are aggregation platforms that stitch together all the edge devices and consolidate and optimize data traffic over a distance of approximately 100M. These devices are often located at the building or small campus level and can analyze, filter, combine and prioritize communications, sometimes with artificial intelligence (AI) cooperation.
Intermediate platforms are positioned on the level below massive central data centers (core) to deliver quicker responses. These responses are often algorithm-based selected by and periodically updated from the core. These platforms provide the real-time control needed for the platforms closer to the edge. Data analysis and tracking from these platforms can provide value and new income streams to the platform providers. Cost savings from operations such as predictive maintenance, material tracking and routing, system management, and data traffic load balancing can be passed on to the user (perhaps for a subscription fee or percentage of the savings).
Big data operations are run on core data-center platforms. These massive data processing and storage facilities hold years’ worth of historical operations and complex machine-learning algorithms that provide the optimization filters and processes to program intermediate platforms for quick responses─and added value to customers.
Evaluating Different Types of Data Center Platforms
At each level of the 5G hierarchy, different requirements and trade-offs are presented to the designer:
- Small platforms are the most cost-, footprint- and power-constrained elements of the system. They need to be easy to install and require a mid-range lifetime because several are needed for each aggregation platform. MCU-based implementations can be used here since they provide the low cost, low power, and small footprint required while making some minor sacrifices regarding flexibility and processing power.
- Aggregation platforms require significant flexibility, processing power, intermediate storage, and security. Field-programmable gate array-based implementations can provide a maximum amount of flexibility and processing power since the underlying hardware can be reprogrammed as needed for new protocols, new AI algorithms, or new value adds to customers. FPGAs can also be easily scaled, allowing providers to create different product tiers with varying flexibility and processing power levels at different price and value points.
- Intermediate platforms require the highest level of raw processing power, security, and flexibility. Cost, power consumption, and footprint are acceptable sacrifices. At this level, a hybrid combination of memory protection units (for raw processing power of common operations) and FPGAs (for flexibility and adaptability) will be the optimal implementation approach. The FPGA can even be reprogrammed in real time on an as-needed basis to respond to the increased need for important features such as video processing, encryption, decryption, searching, and filtering. AI algorithms can even predict such needs by analyzing clues they find in key metrics such as occupancy, traffic patterns, weather, etc.
How an Aggregation Platform Works
Let’s take a look at a mid-range featured aggregation platform to see how key features are implemented within the device. Utilizing an FPGA with an on-chip microcontroller allows the device to begin operation from the MCU, providing a known secure starting point for boot and secure updates. This root of trust uses robust and protected encryption, decryption, and security key storage operations to thwart hackers and viruses from taking over the system. The MCU can also handle standard operations such as communications, packet processing, video processing, compression, and storage efficiency. The on-chip FPGA hardware can be used for the less common but processing-power-intensive operations to keep the MCU freed up for common operations that must be completed in a timely manner.
The FPGA can be used for digital filtering, image processing, image recognition, and similar special operations, perhaps in conjunction with AI and machine-learning algorithms to predict and program the hardware on an as-needed basis. Over time, new algorithms could be identified, created and downloaded from intermediate and core platforms to further optimize performance and to create new revenue streams for the platform provider─and cost savings for the customer.
Conclusion
5G will mean bringing together cloud, core and the edge. However, it’s important that each of these runs on─or has access to─the right types of infrastructure. A core will always be needed, even with the importance of edge computing for 5G, but as 5G connectivity proliferates and the number of 5G-enabled devices continues to grow, so will the need for small and intermediate distributed data center platforms for those devices. Computing from the core all the way down to the edge will be the hallmark of the 5G-enabled world. It will be a world in which thermostats and refrigerators to planes, trains and automobiles are connected to the same cellular network as cell phones.