Vous êtes ici : Accueil » Kiosque » Annonce

Identification

Identifiant: 
Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?

Annonce

13 novembre 2018

[Master] Modeling of hardware-software partitionning for 5G CRAN architecture


Catégorie : Stagiaire


Contact : Matthieu Gautier (matthieu.gautier@irisa.fr) , Robin Gerzaguet (robin.gerzaguet@irisa.fr)

Keyword: FPGA, SDR, RFNoC, C-RAN

Location: Granit team, IRISA Laboratory, Lannion

Context

Cloud Radio Access Networks (C-RANs) have been introduced to reduce the cost of base station deployment and management [1][2]. C-RAN approach is often viewed as an architectural evolution of the distributed base station concept, which allows separating the physical location of the remote radio head (RRH) from the digital baseband unit (BBU). This helps to facilitate coordination between potentially interfering base station systems but also it allows sharing between resources of different base stations. C-RAN approach has the potential to decrease cost, energy and power consumption compared to traditional approaches and is therefore expected to be widely adopted by 5G systems. However, one of the main challenges is to determine how to separate the processing between the base station and the cloud. The RRH/BU partitioning may be different from a RRH to another. Moreover, this functional partitioning can be redistributed among time, depending on several criteria related to energy consumption, network congestion, latency and throughput. It forces to envision adaptive base station capable to cope with various congurations. It also spotlights the importance of the so-called orchestrator which is in charge of the dynamic RRH-BBU split.

 

Internship goals

In order to evaluate the efficiency of the RRH/BBU partitioning, accurate models for energy consumption, throughput and latency both at the RRH and the BBU must be proposed. Moreover, different hardware targets of the process must be considered. These models will help the orchestrator, whom specifies the partitioning, for rapid prototyping. A state of the art analysis on how these models are built will be firstly performed. This analysis will aim to propose several key performance indicators for the orchestrator, linked to final desired conguration. Based on this analysis, several models will be considered, which correspond to different processing mappings. The first corresponds to a full-cloud approach (no processing in the RRH). The second one will be linked to a full base station system (no processing in the cloud). In order to take advantage from RRH/BBU partitioning, an intermediate partitioning with few processing in the RRH (e.g. synchronization) will finally be modeled.

From this model, a reconfigurable orchestrator will be proposed, based on the defined KPI and the different energy models proposed. This orchestrator will efficiently explore the design space thanks to a microbenchmarking of the different components (Hardware and Software). A demonstration on a Software Defined Radio platform (based on USRP e310 devices [3]) will conclude the internship.

The internship will focus on the following steps:

- Analysing state of the art on energy partitioning in CRAN architecture,

- Proposing of energy/throughput models for different RRH/BBU partitioning,

- Microbenchmarking the considered hardware and software blocks and proposing different functional splits between RRH and BBU,

- Validating network behavior with the various functional splits with USRP-e310 cores,

- Evaluating performance (in terms of throughput, latency and energy consumption) for different RRH-BBU split models.

References

[1] A. Checko and et al., "Cloud RAN for Mobile Networks; A Technology Overview," IEEE Communications Surveys Tutorials , 2015.

[2] S. Barbarossa, S. Sardellitti, and P. D. Lorenzo, "Communicating While Computing: Distributed mobile cloud computing over 5G heterogeneous networks," IEEE Signal Processing Magazine , 2014.

[3] Ettus Research, "Universal Software radio platform (USRP)," 2017. https://www.ettus.com/product/details/E310-KIT.

 

Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2018.