Cache-aided Folding for Reducing VoD Loads in Networks

Project Information

LIGHT

Grant agreement ID: 101101031

DOI

10.3030/101101031

Project closed

EC signature date 16 January 2023

Start date 1 March 2023

End date 31 August 2024

Funded under

• European Research Council (ERC)

Total cost

No data

EU contribution

€ 150 000,00

Coordinated by

EURECOM GIE
France

Periodic Reporting for period 1 - LIGHT (Cache-aided Folding for Reducing VoD Loads in Networks)

Reporting period: 2023-03-01 to 2024-08-31

Summary of the context and overall objectives of the project

The Problem: The astronomical costs of communicating VoD data arise from both wired and wireless networks, which face pressure to accommodate a projected tenfold increase in data traffic over the next five years. VoD applications now account for over 70% of this traffic, driving up costs for wireless providers as they purchase expensive bandwidth and infrastructure to maintain high Quality of Service (QoS). Wired network providers must constantly overprovision their networks, further escalating costs. Content providers like Netflix and Amazon also incur significant fees per content volume to deliver their services.
The Setting: Networks utilize caches, but their current application is inefficient. Content delivery networks employ numerous caches to push content closer to users, yet this approach only marginally reduces data volumes. For instance, if a cache holds 10% of a VoD provider's catalog, only 10% of traffic will decrease, leaving providers responsible for the remaining 90%. Thus, simply pushing content closer is not effective.
The LIGHT Software Solution: LIGHT proposes using caches to fold and shrink VoD data instead of just pushing it closer to users. In the same network setup, we envision a software solution installed at both the data source and the caches that reduces traffic volumes by an additional factor of three, without additional compression or QoS loss. Our previous work has shown that properly cached data at receivers can significantly enhance the performance of modern wired and MU-MIMO wireless networks. With further algorithmic development, LIGHT aims to leverage existing caches to condense data streams, thereby substantially lowering the volume of VoD data communicated without altering the end-user experience. This approach could be implemented across various scenarios, such as streaming services, CDNs, and mobile or satellite internet providers, where VoD data transmission costs are high. Ultimately, LIGHT has the potential to reduce VoD transmission volumes, cut costs, and enhance streaming speeds.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

In this ERC-PoC project, we developed two TRL3 demonstrators for wireless and wired networks to validate our technology. The wireless demonstrator uses a 5G-compliant link-level simulator (LLS) to assess the Physical Downlink Shared Channel (PDSCH). Built with MATLAB and C code from the Open-Air-Interface (OAI) platform, it features a gNB (base station) with L transmit radio units (TXRU) serving K single-antenna UEs via MU-MIMO. The gNB uses Zero-Forcing (ZF) beamforming to mitigate interference, adhering to 3GPP standards for LDPC and CRC encoding, with coherent detection at the receiver. With LIGHT technology activated, UEs subtract interference by encoding raw bits with transmission parameters and the RNTI (Radio Network Temporary Identifier) of interfering UEs.
The wired demo implements LIGHT's higher-level components (application to network layer) over unicast (TCP) and multicast (UDP) on a Local Area Network (LAN). Written in Python and packaged as Docker containers, it includes server and receiver functionalities for video streaming and memory management. Unicast fills caches, while both unicast and multicast support video delivery. Tested with 2 servers and 12 receivers (laptops and Raspberry Pi devices), the demo achieves over a 3-fold reduction in video delivery data volumes.
At the same time, we have defined the product Requirements Specifications (PRS), to outline the LIGHT design problem, detailing high-level requirements for various applications. This helped establish clear expectations, facilitating consistent work plans and viable contracts with clients. We have also successfully designed and packaged the LIGHT software, including algorithmic modules for real-time data consolidation, cache placement, and the folding algorithm for multiple use cases. In addition, we have provided various tests in diverse settings, including realistic emulations involving OAI modules. Testing focuses on network topology in wired settings and channel randomness in wireless environments. The main target scenarios included basic wired and wireless networks.
We have additionally explored three business use cases for our proposed solution: wired networks (e.g. CDNs), wireless networks (mobile/5G or Wi-Fi), and satellite networks. We focused on the technical feasibility and business value of wired and wireless networks, creating presentations to communicate the technology's benefits and gather insights from potential users and investors.
Part of the market research, we also have participation in industry events like the Mobile World Congress (MWC) 2024 and the RIPE88 meeting allowed us to network with key stakeholders and identify potential partners. We established connections with experts in CDN, mobile network, and satellite communication sectors, engaging in discussions with Cloudflare and Nokia, both showing interest in our technology.
For funding acquisition, we are pursuing EU grants through the European Innovation Council (EIC), aiming to develop TRL6 demonstrators for both wired and wireless environments. Additionally, we prepared a business model canvas and a pitch deck for private funding applications to venture capitalists.

We have successfully adapted our original algorithm for MU-MIMO settings, extending its application to various wireless and wired networks. Key achieved milestones include:
• Achieving TRL3 verification with a 300-400% capacity increase compared to optimized state-of-the-art systems, alongside 3-4 times bandwidth reduction.
• Demonstrating a 4x reduction in required antennas (reducing RF chains), substantial energy savings for base stations, and a 4x decrease in channel-feedback overhead.
• Notably, we achieved a two-fold reduction in data volumes while ensuring compatibility with security protocols in a two-server wired network.
• Secured intellectual property for the core algorithm applicable to both wireless and wired networks.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

Potential Impact: LIGHT could significantly lower infrastructure costs in Europe, reducing expenditures from $100 billion to much less by enabling the use of cheaper base stations (BSs) with fewer RF chains—averaging $30,000 for units with 4 chains compared to $250,000 for those with 64 or more. With around 356,000 5G base stations deployed in the EU, this technology promises substantial savings. LIGHT also enhances energy efficiency for small and large BSs, improving transmission power and reducing amplifier power. By streamlining operations and lowering computational loads, it lowers initial capital expenditures and offers a return on investment within 3 to 5 years, potentially increasing revenue by 20-40% from doubling network capacity while cutting costs per bit by 30-50%.

Next Steps: To advance the LIGHT project, we aim to move from TRL3 to TRL4-5 and ultimately to TRL6 by developing a TRL6 emulator for the complete 5G stack using the OpenAir Interface at EURECOM. Key tasks include software development, analyzing cross-layer interactions, calibrating implementations for BSs and user equipment, ensuring compliance with 5G standards, and refining algorithms.

Additionally, we will ensure LIGHT's compatibility with current and future telecommunications standards (5G and 6G) by engaging with industry leaders and the 3GPP community for feedback, support, and proposals.

Base station with L=16 RF chains, ZF: rate gains over MU-MIMO