Coded Caching for Improved Data Rates and Networking in 6G

When we looked at Achieving the Terabit/s Goal in 6G Broadband Connectivity in a blog post, it referenced a research paper available here. One of the topics that the paper discussed was CC.

Coded caching (CC) was originally proposed as a way to increase the data rate with the help of cache memories available throughout the network. It enables a global caching gain, proportional to the total cache size of all users in the network, to be achieved in addition to the local caching gain at each user. This additional gain is achieved by multicasting carefully created codewords to various user groups, so that each codeword contains useful data for every user in the target group.

Technically, if K is the user count, M denotes the cache size at each user and N represents the number of files in the library, using CC the required data size to be transmitted over the broadcast link can be reduced by a factor of 1 + t (equivalently, the data rate can be increased by a factor of 1 + t), where t = KM/N is called the CC gain.

Interestingly, the CC gain is not only achievable in multiantenna communications, but is also additive with the spatial multiplexing gain of using multiple antennas. In fact, caching non-overlapping file fragments at user terminals provides multicasting opportunities, as multiple users can be potentially served with a single multicast message. Therefore, the additive gain of CC in multi-antenna communications can be achieved through multicast beamforming of multiple parallel (partially overlapping) CC codewords to larger sets of users, while removing (or suppressing) intercodeword interference with the help of carefully designed beamforming vectors. The generalized multicast beamforming design enables innovative, flexible resource allocation schemes for CC. Depending on the spatial degrees of freedom and the available power resources, a varying number of multicast messages can be transmitted in parallel to distinct subsets of users.

The multi-antenna CC structure also provides more flexibility in reducing subpacketization, defined as the number of smaller files each file should be split into, for the CC structure to work properly. An exponentially growing subpacketization requirement is known to be a major problem in implementing original single- and multi-antenna CC schemes. However, recently it is shown that in multi-antenna CC, especially when the spatial multiplexing gain is larger than the CC gain, linear or near-linear subpacketization growth is possible through well-defined algorithms. Overall, the nice implementation possibility of CC in multi-antenna setups makes it a desirable option to be implemented in future wireless networks, where MIMO techniques are considered to be a core part.

The number of multimedia applications benefiting from CC is expected to grow in the future. One envisioned scenario assumes an extended reality or hyper-reality environment (e.g., educational, industrial, gaming, defense, social networking), as depicted in Fig. 30. A large group of users is submerged in a network-based immersive application, which runs on high-end eye-wear that requires heavy multimedia traffic and is bound to guarantee a well-defined quality-ofexperience level for every user in the operating theatre. The users are scattered across the area covered by the application and can move freely, and their streamed data is unique and highly location- and time-dependent. Notably, a large part of the rich multimedia content for rendering a certain viewpoint is common among the users. This offers the opportunity for efficient use of pooled memory resources through intelligent cache placement and multicast content delivery mechanisms. In such a scenario, the possibility of caching on the user devices and computation offloading onto the network edge could potentially deliver high-throughput, low-latency traffic, while ensuring its stability and reliability for a truly immersive experience. The fact that modern mobile devices are continuing to increase their storage capacity (which is one of the cheapest network resources) makes CC especially beneficial given the uniqueness of this use case, where the popularity of limited and location-dependent content becomes much higher than in any traditional network.

You can continue reading here, jump to page 37. A slightly older video presentation which is a part of the project Arctic 5G Test Network is embedded below and will help understand the topic better.

Free 6G Training

Coded Caching for Improved Data Rates and Networking in 6G

Comments

Post a Comment