Title :
DPDK and MKL; Enabling technologies for near deterministic cloud-based signal processing
Author :
Alizadeh, Roya ; Belanger, Normand ; Savaria, Yvon ; Frigon, Jean-Francois
Author_Institution :
Dept. of Electr. & Comput. Eng., Ecole Polytech. de Montreal, Montreal, QC, Canada
Abstract :
This work explores different means to implement software defined cellular radio over the cloud, which sets stringent real-time requirements. DPDK and MKL are two technologies proposed by Intel that can be used for that purpose. In this paper, we explore those technologies and using operating system isolation to reduce the computation time variability for LTE processes. It is found that MKL offers excellent performance, but as it is not scalable by itself over the cloud, combining it with DPDK is a very promising alternative. DPDK allows improving performance, memory management and makes MKL scalable.
Keywords :
Long Term Evolution; cellular radio; cloud computing; operating system kernels; signal processing; software radio; telecommunication computing; DPDK; LTE process; Long Term Evolution; MKL; computation time variability reduce; data plane development kit; math Kernel library; memory management; near deterministic cloud-based signal processing; operating system isolation; software defined cellular radio; Libraries; Long Term Evolution; Memory management; Random access memory; Signal processing; Software; CPU core isolation; Data Plane Development Kit (DPDK); Fast Fourier Transform (FFT); Long Term Evolution (LTE); Math Kernel Library (MKL); deterministic computations;
Conference_Titel :
New Circuits and Systems Conference (NEWCAS), 2015 IEEE 13th International
Conference_Location :
Grenoble
DOI :
10.1109/NEWCAS.2015.7182086