Page 19 - Kaleidoscope Academic Conference Proceedings 2021
P. 19
6G TECHNOLOGIES FOR MOBILE CONNECTED INTELLIGENCE
Geng Wu
Intel Fellow, Intel Corporation, USA
With the introduction of ultra-reliable and low-latency communications, 5G-enabled massive machine type
communications beyond traditional smartphone applications and services. To achieve the required network
performance and service scalability, 5G systems have adopted several technologies such as edge computing,
network slicing and service-based architecture, powered by advanced virtualization technologies. Building on
the technology foundations established in 5G, 6G is expected to bridge the physical and virtual worlds and
bring technical improvements to energy efficiency, security, and network resilience. More specifically, 6G
systems will integrate computing, communications, and data storage and access functions. 6G network
architecture is expected to enable scalable distribution of computing and intelligence workloads, with the
support of micro-services across devices, network edge and cloud. A next generation programmable optical
and wireless transport network infrastructure will be developed and deployed to support 6G systems.
There are several important trends that drive the development of 6G technologies. The first and foremost is
the shifting nature of data. In addition to the unprecedented growth in total data volume throughout the network,
we see a much faster growth at the network edge, driven by the proliferation of machine learning and artificial
intelligence. This new data tends to be distributed and mobile, of diverse numeric format, precision, and quality.
Due to the local nature of sensing, they are best processed near or at the source due to their large number of
sources, massive volume, and the increasing need for privacy and security.
Another important driving force is the significant increase in wireless link peak data rates. For 6G, we can
expect a maximum of 200 Gbps over the downline (from base station to mobile device) and 100 Gbps over the
uplink (from mobile device to base station). Such a data rate increase is possible thanks to the more advanced
radio link design and signal processing technology, as well as the addition of new spectrums including the
terahertz bands. These significant increases in data rate and the continuing reduction in latency give 6G
systems an unprecedented capability to support wide-area mobile distributed computing. In fact, the target data
rates of 6G radio links are approaching that of many of interconnecting technologies we use today in our
computing systems.
Computing workload in the 6G era is also changing. With the proliferation of machine learning and artificial
intelligence, computing workloads are increasingly distributed throughout a wide area, with stringent
requirements on latency, energy efficiency and user privacy. Taking the federated learning use case as an
example, training data from local sensors are first processed at the device to preserve user privacy and to avoid
energy costs associated with sending the training data over the radio link; the infrastructure cloud further
processes the models it has received from many devices to produce a complete and robust final model. To
enable these types of mobile computing, it is important for 6G to support computing scaling out from network
infrastructure to mobile devices. This scaling out happens at all levels of computing, including micro-services
at the service level and disaggregated computing at the computing resource level. It often involves
heterogenous forms of computing fabrics and network architecture.
The above technology trends lead to the network evolution from the communication-centric 4G/5G to the
compute-centric 6G, where computing, communication, and data storage/access are expected to come together.
We expect major structural changes in the designs of air interface, network architecture, and signal processing
for cloud-native and AI-native 6G systems. To be more specific, we envision the introduction of a new
Compute Plane and a new Data Plane in 6G systems from the device to the network; we also expect 6G air
interface to be designed with built-in native AI capabilities. There is also a need for a major upgrade of
networking fabric and transport for 10x-100x platform performance improvement.
– xv –