A Distributed Real-Time Programmable Scheduler Architecture For Cellular Networks
|
|
University of Sydney: Wibowo Hardjawana (wibowo.hardjawana@sydney.edu.au), Branka Vucetic (branka.vucetic@sydney.edu.au), Dawei Tan, Zhouyou Gu, Wenhao Zhang Telstra: Simon Lumb, David McKechnie and Todd Essery |
Abstract
Two types of cellular network scheduler architectures have been proposed in the open literature: 1) a distributed scheduler, as shown in Fig. a, with non-programmable logic for network operators due to tight control by the evolved NodeB (eNodeB) vendor at the edge base station (BS), and 2) a centralized scheduler, as shown in Fig. b, with only non-real-time scheduling logic that is programmable. We propose a new distributed real-time programmable scheduler architecture, as shown in Fig. c. We refer to a real-time scheduler as one with logic running every transmission time interval (TTI). The scheduling logic is written independently of the underlying eNodeB software and executed in real-time at the edge BS with the help of a scheduler agent. The proposed architecture is validated in an over-the-air environment with commercial long-term evolution (LTE) devices and 3rd Generation Partnership Project (3GPP) standards-compliant setup.
Fig: (a) Distributed non-programmable; (b) centralized programmable; and (c) distributed programmable scheduler architectures.
Contributions
Firstly, we believe this is the first architecture that allows network operators to rewrite scheduling logic at the controller and that uses a scheduler agent at the edge BS to execute that logic in real-time every TTI, as shown in Fig. c. To date, the most advanced centralized programmable architecture, as shown in our experiments, struggles to work in real-time due to a communication latency between the scheduler and eNodeB that exceeds multiple TTIs. Secondly, the agent incorporates new parameters on radio resource frame configurations when interfacing the scheduler to the eNodeB. These parameters have not been used in any scheduler interface standards such as FAPI. In the current FAPI implementation, the scheduler needs to calculate these parameters, resulting in more information exchange between scheduler and eNodeB and a complex scheduler design. Inclusion of these parameters into FAPI are recommended. Thirdly, we develop a prototype for the above architecture and provide examples how to create scheduling functions to control traffic priority and to slice radio resources for the downlink logical data channels at the controller in over-the-air environments with commercial LTE devices and an LTE standards-compliant setup, referred to as an experimental LTE network.
Publication and Presentation
Z. Gu, W. Hardjawana, W. Zhang, D. Tan, B. Vucetic, S. Lumb, D. McKechnie, and T. Essery, A distributed real-Time programmable scheduler architecture for cellular networks,” submitted to ACM Computer Communication Review, 2019.
S. Lumb et al.,“Open Programmable Scheduler,” presented in O-RAN Working Group 8 Meeting during MWC 2019
(download full paper)
A Real-Time Vendor-Neutral Programmable Scheduler Architecture for Cellular Networks
|
|
University of Sydney: Wibowo Hardjawana (wibowo.hardjawana@sydney.edu.au), Branka Vucetic (branka.vucetic@sydney.edu.au), Wenhao Zhang, Zhouyou Gu Telstra: Simon Lumb, David McKechnie and Todd Essery |
Abstract
The current Downlink Shared Channel (DLSCH) resource scheduler for cellular networks has the following features: 1) it is integrated with an evolved NodeB (eNB) and 2) uses proprietary interfaces. The first causes a temporary outage whenever the scheduler logic is reprogrammed to accommodate traffic profiles that have different requirements, while the latter prevents multi-vendor interoperability. In this paper, we propose a real-time vendor-neutral programmable DLSCH scheduler architecture. The scheduler and eNB are separated into two binary files that communicate via an agent. The agent uses standard interfaces to interpret information from/to different eNB vendors in real time. The proposed architecture is implemented on two open source 3rd Generation Partnership Project standardcompliant eNB stacks from the OAI and SRS. Experimental results show that the proposed architecture addresses the real time and proprietary challenges mentioned above.
Fig: RTVN programmable scheduler architecture.
Contributions
The first contribution here is that the DLSCH scheduler logic at an eNB will become real-time programmable. This is achieved by separating the DLSCH scheduler from the eNB as a separate binary file, referred to as a shared library. The remaining eNB functions are then run in another binary file, referred to as ReNB. As a result of the separation, the logic of the DLSCH can be changed and compiled individually and in real time, without interrupting the operation of the eNB. This is in contrast with [1] and [2], where any change made to the scheduler logic will result in a temporary outage at the eNB. The second contribution is a vendor-neutral DLSCH scheduler that works with the ReNB vendor. This is achieved by standardising the interface parameters exchanged between the scheduler and ReNB, referred to as the global variable (GV). This is in contrast with [1] and [2], which employ proprietary interfaces. The third contribution is the development of a vendor-neutral controller that can generate the shared library containing the DLSCH scheduler logic. The scheduler logic is written by operators and complied as a shared library at the controller, using the GV as its interface parameters.
Publication and Presentation
Wenhao Zhang, Zhouyou Gu, Wibowo Hardjawana, Branka Vucetic, Simon Lumb, David McKechnie and Todd Essery, A Real-Time Vendor-Neutral Programmable Scheduler Architecture for Cellular Networks, IEEE WCNC 2020
(click to download full paper)