Mérouane Debbah, ERC 2017 grant holder from CentraleSupélec, presents his project CacheMire
Massive data to enhance the performance of 5G systems
Throughout the ERC MORE project (Advanced Mathematical Tools for Complex Network Engineering), the team of Mérouane Debbah from LANEAS group has developed, among other things, the concept of “caching on the Edge” in order to improve the performances of the 5G systems. 5G systems will deploy a massive number of antennas, either centrally (Massive MIMO) or in a decentralized manner (Small cells) in order to provide the 10 gigabit experience. This deployment will require an expensive backhaul (wireless or wired cables connecting the antenna to the heart of the network) which limits in general the network's capacity. Typically, the factor limiting the wireless broadband rate in the house is not the Wi-Fi but the ADSL connections. On another level, these new network architectures increase the end to end latency of the client’s experience, which is not compatible with the future use of 5G, such as virtual reality or the connected car. At the beginning of the ERC grant MORE in 2012, the team proposed (among other things) the idea to exploit the massive data of networks, in line with the “Big Data for the Telecom” approach, in order to make increase the quality of experience. That was against the ideas of the telecommunication community who was seeing the increased data flood as a curse rather than an opportunity. The approach was based on machine learning technique that could extract useful information to make the network predictable, pro-active and especially reduce the burden on the backhaul.
CacheMire (Wireless Edge Caching Platform) aims to embed the antennas with storage units (“Edge caching”) and computing units (“Edge computing”) in order to pre-store the contents of the users before their actual requests using the precise analysis of the data passing through the network. The results show very important gains in terms of rate but also of latency and energy efficiency of the network. In that respect, the work of the members of Mérouane Debbah’s team, in particular, Ejder Bastug, has received numerous prices since 2012.
From a scientific perspective, the development of the “Caching on the Edge” concept will help the implementation of new mathematical and engineering tools both at the learning level (Mean Field Games, Echo state Networks, etc.) and on the engineering level of networks and their performance (Random Matrices and Stochastic Geometry).
Interview with Mérouane Debbah, CacheMire project coordinator
Could you present your research background?
I am a specialist of random matrix and game theory with applications to wireless networks. I work at the interface of information theory, signal processing and telecommunications.
You have already received an ERC grant. Could you explain what the difference between these 2 ERC funding grants is?
The ERC project MORE has allowed me to develop the theoretical abstraction and the mathematical tools necessary for the engineering of the concept. Nowadays, we are facing complex problems with millions of parameters to configure. This is not possible with existing tools, which do not scale. With the CacheMire project, we have developed new mathematical tools allowing to better understand things and above all to configure the parameters in the asymptomatic regime.
What were the reasons for applying for Proof of Concept grant?
The Proof of Concept CacheMire allows to develop a prototype to verify and validate the theoretical concepts introduced during the first ERC.
How this funding helped you in your research?
The Proof of Concept is a unique opportunity to pass from theory to practice and to compare the theoretical hypotheses with the reality in order to improve the model.
What advice would you give to those who want to apply for the Proof of Concept grant?
From my point of view, the Proof of Concept is a natural outcome of the ERC and I strongly encourage all researchers having the ERC Grant to proceed with an application. This allows to better disseminate the results of their ERC but also to be able to compare the abstractions and the hypotheses made (which are necessary to consider the disruptive innovations) with the network reality.