Enabling Ultra-Low Delay Teleorchestras using Software Defined Networking

Ultra-low delay sensitive applications can afford delay only at the level of msec. An example of this application class are the Networked Music Performance (NMP) systems that describe a live music performance by geographically separate musicians over the Internet. The present work proposes a novel architecture for NMP systems, where the key-innovation is the close collaboration between the network and the application. Using SDN principles, the applications are enabled to adapt their internal audio signal processing, in order to cope with network delay increase. Thus, affordable end-to-end delay is provided to NMP users, even under considerable network congestion.



There are no comments yet.



A Novel QoE-Aware SDN-enabled, NFV-based Management Architecture for Future Multimedia Applications on 5G Systems

This paper proposes a novel QoE-aware SDN enabled NFV architecture for c...

A Deep Learning Approach for Low-Latency Packet Loss Concealment of Audio Signals in Networked Music Performance Applications

Networked Music Performance (NMP) is envisioned as a potential game chan...

Blockchain-enabled Authentication Handover with Efficient Privacy Protection in SDN-based 5G Networks

5G mobile networks provide additional benefits in terms of lower latency...

Delay Analysis of Base Station Flow Table in SDN-enabled Radio Access Networks

Future generation wireless networks are designed with extremely low dela...

Delay-Efficient and Reliable Data Relaying in Ultra Dense Networks using Rateless Codes

We investigate the problem of delay-efficient and reliable data delivery...

Centralized Congestion Control and Scheduling in a Datacenter

We consider the problem of designing a packet-level congestion control a...

Time Delay Estimation of Traffic Congestion Propagation based on Transfer Entropy

Considering how congestion will propagate in the near future, understand...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

1. Introduction

Networked Music Performance (NMP) systems, describe the process where musicians located in different places perform synchronized via the Internet (Lazzaro and Wawrzynek, [n. d.]). NMPs belong to ultra-low delay sensitive applications due to the latency requirements they have. In NMP services the maximum affordable delay between the transmitted and the finally reproduced signal should be up to 25 ms. This constraint is denoted as the Ensemble Performance Threshold (EPT) (Schuett, 2002).

There are many factors that affect end-to-end delay in NMP systems, which are grouped in two categories: network delay and audio processing delay. Network delay expresses the delay due to data transmission via the network, attributed to physical propagation and the network operating state. On the other hand, the audio processing delay is introduced by the audio capturing, processing and encoding methods, for each transmitter/receiver pair. Due to the delay that audio encoders yield (about 100 ms), they are avoided in NMP services; instead, raw audio is preferred (Akoumianakis et al., [n. d.]; Goto et al., 1910).

In this work, our goal is to endow NMP systems with resistance against traffic congestion and link failure cases. To make this feasible, our system introduces collaboration between the NMP application and the network. In more detail, the proposed architecture exploits the flexibility for dynamic network reconfiguration and global view of the network condition that Software Defined Networking (SDN) offers in order to achieve the acceptable latency for NMP (Durner et al., [n. d.]; Gorlatch et al., [n. d.]; Kobayashi et al., 2014). The SDN controller, that orchestrates the network, is responsible for the data path setup that will carry the audio between each transmitter/receiver pair. Additionally, the controller is responsible for rerouting audio flows in case of congested paths. In cases where the whole network is congested, offering no feasible paths for NMP, the SDN controller communicates with the NMP application, at both the transmitter and receiver sides, in order to switch to another audio configuration set, thus decreasing the audio processing delay and compensating the increased network latency. Raw audio is used, instead of encoders (Maribondo and Fernandes, [n. d.]; Wabnik et al., 2009; Valin et al., 2016), due to their latency constraints. The proposed architecture, avoids network bandwidth waste, since each node receives a single version of the initial signal, instead of different encoded versions of the same signal (Akoumianakis et al., [n. d.]; Alexandraki et al., 2008; Baltas and Xylomenos, 2014; Carôt and Werner, 2007; Carôt et al., 2006, 2007; Xiaoyuan Gu et al., 2005). The proposed Software-Defined NMP solution can be also applied in more generic use cases encountered in the real Internet, e.g., considering the domains (Autonomous Systems) between the sender and the receiver as “big switches”, abstracted with the use of classic tunneling mechanisms (Kotronis et al., 2016, 2014). This approach is scalable since it requires only a number of points of presence anywhere in the network where paths can be programmatically set up and monitored either in the form of overlay tunnels or physical links.

Similar approaches try to guarantee QoS via applying certain policies on the traffic queues of the switches/routers (Kumar et al., 2013; Sharma et al., 2014a; Durner et al., 2015; Sieber et al., 2015; Egilmez and Tekalp, 2014; Adami et al., 2015), based on different criteria such as source and destination IP address, transport protocol or generally applying either Type of Service (TOS) matching (Sieber et al., 2015; Egilmez and Tekalp, 2014; Adami et al., 2015; Sharma et al., 2014b) or DiffServ approach (Tomovic et al., 2014). On the other hand, we apply smart routing in the network based on real-time measurements instead of incorporating metrics that reflect the difference between the transmitted and the required by the user audio quality  (Mu et al., 2016; Koumaras et al., 2016) combined with the adjustment of application-tunable parameters, in order to deal with sudden increases in network delay. We evaluated our system in an emulation environment, demonstrating the advantages of the proposed architecture.

2. The SDN Teleorchestra System

As described in Section 1, end to end delay in NMPs is the summary of network and audio processing delay. In mathematical form, end to end delay in NMPs for similar user audio equipment can be modeled as:


where represents end-to-end delay, shows the delay due to sound-card in transmitter and receiver and expresses the network delay. Delay due to audio processing is alternatively called as blocking delay. Equation (2) describes the blocking delay evaluation:


In equation (2), frame size denotes the size of audio packets that a user sound-card can process per hardware clock tick, and the sampling rate represents the number of samples the sound-card acquires per second. Finally, is a constant delay that is due to the sound-card’s hardware quality. For blocking delay decrease, the fraction between frame size and sampling rate should be also decreased.

The proposed architecture consists of three components: a SIP, an SDN and a Network Monitoring service as shown in Fig. 1

. The SIP service initially measures the audio performance of each user for different audio settings and stores this information. Additionally, it classifies each user as premium or regular based on the privileges that he has. Premium users do not accept low quality transmissions compared with regular users. This category contains musicians that require high quality audio or users (audience) that may have paid for low latency according to SLAs. SIP triggers the application when audio modification is required

(Nam et al., 2014; Nurmela, 2007; Ali et al., 2013; Sinnreich and Johnston, 2006; Camarillo et al., 2003).

The SDN service describes the controller functionality that installs paths for audio transmission. The number of paths towards each user and the path characteristics are defined by the user classification. Each path consists of OVS switches that are instructed by SDN service via the OpenFlow protocol (Akyildiz et al., 2014; McKeown et al., 2008). Also, the SDN service reroutes audio flows when a better path is available. Finally, the Network Monitoring service measures network delay for all paths between users by periodic UDP packet transmission between them.

Figure 1. The proposed system architecture.

When a new user participates in our system, the SIP service creates an audio profile of the user for various audio parameter combinations. Moreover, it classifies the user as either premium or regular as discussed. Based on the classification results, the SDN service assigns a single path from the audio source to the user and audio transmission starts, and keeps a tunable number of backup paths at the ready. The Network Monitoring service continuously monitors all paths and the SDN service reroutes audio flows to a path that surpasses the active one in performance. If all paths are congested resulting in over-EPT end to end delay, the SIP service contacts the application side in order to switch to an audio mode that has less blocking delay. When the network recovers from congestion, the application returns to the previous audio configuration.

3. System Evaluation

To evaluate our system, we used real live audio, and Mininet  (Lantz et al., [n. d.]; Sharma and Sood, 2014) to emulate a realistic network setup. The topology is shown in Fig. 1. POX  (Prete et al., [n. d.]; Kaur et al., 2014; Bagewadi and Babu, 2014; Sukhveer et al., [n. d.]; Silvan and Christos, [n. d.]) is selected as the SDN controller. We used Netem traffic control tool in order to increase delay in each path. To avoid redundant reroutes, we used a threshold of at least 2 ms lower latency as the criterion for taking rerouting decisions among available paths. The results of our experimental evaluation are depicted in Fig. 2. denotes sampling rate and the frame size. The proposed method can also be applied in conventional Internet topologies enabling guaranteed end-to-end delay via inter-domain routing  (Liaskos et al., 2016a; Gkounis et al., 2016; Liaskos et al., 2016b; Liaskos, 2015).

Assume User B declares interest for receiving audio from User A. There are three paths between User A and User B in the test topology. At the beginning, the SDN service assigns path for audio transmission with equal to kHz and set to samples. As we increase the network delay in the path, the SDN service reroutes audio flows to paths (at s) and (at s). At s, the SIP service informs the application to switch to an audio mode with lower blocking delay ( equal to KHz and unaltered ). Assuming that the network delay increases further, and in order to keep end-to-end delay below EPT, at s, the application modifies to samples via interaction with the SIP service. This process takes place during the experimental setup for various audio modes. When the network delay is very high, our system results in best effort delay (after s), given the available audio modes and network condition. Comparing our method with the case that application and network could not interact, we achieve delay improvement by in end-to-end delay.

Figure 2. End-to-end delay evaluation.
This work has been funded by the European Research Council Grant Agreement no. 338402.


  • (1)
  • Adami et al. (2015) Davide Adami, Lisa Donatini, Stefano Giordano, and Michele Pagano. 2015. A network control application enabling Software-Defined Quality of Service. IEEE, 6074–6079.
  • Akoumianakis et al. ([n. d.]) D. Akoumianakis et al. [n. d.]. The MusiNet project. In IEEE IISA’14.
  • Akyildiz et al. (2014) Ian F. Akyildiz, Ahyoung Lee, Pu Wang, Min Luo, and Wu Chou. 2014. A roadmap for traffic engineering in SDN-OpenFlow networks. Computer Networks 71 (Oct. 2014), 1–30.
  • Alexandraki et al. (2008) Chrisoula Alexandraki, Panayotis Koutlemanis, Petros Gasteratos, Nikolas Valsamakis, Demosthenes Akoumianakis, Giannis Milolidakis, G Vellis, and D Kotsalis. 2008. Towards the Implementation of a Generic Platform for Networked Music Performance: the DIAMOUSES Approach.. In ICMC. 251–258.
  • Ali et al. (2013) Akbar Ali, Nehal Ahmad, Mohd Sharique Akhtar, and Aditya Srivastava. 2013. Session Initiation Protocol. International Journal of Scientific and Engineering Research 4, 1 (2013), 1–6.
  • Bagewadi and Babu (2014) Abhishek Bagewadi and RM Babu. 2014. Towards an Ethernet Learning Switch and Bandwidth Optimization using POX Controller. International Journal of Advanced Research in Computer and Communication Engineering 3, 7 (2014), 7531–7535.
  • Baltas and Xylomenos (2014) George Baltas and George Xylomenos. 2014. Ultra low delay switching for networked music performance. IEEE, 70–74.
  • Camarillo et al. (2003) G. Camarillo, R. Kantola, and H. Schulzrinne. 2003. Evaluation of transport protocols for the session initiation protocol. IEEE Network 17, 5 (Sept. 2003), 40–46.
  • Carôt et al. (2006) Alexander Carôt, Ulrich Krämer, and Gerald Schuller. 2006. Network music performance (NMP) in narrow band networks. In Audio Engineering Society Convention 120. Audio Engineering Society.
  • Carôt et al. (2007) Alexander Carôt, Pedro Rebelo, and Alain Renaud. 2007. Networked music performance: State of the art. In Audio engineering society conference: 30th international conference: intelligent audio environments. Audio Engineering Society.
  • Carôt and Werner (2007) Alexander Carôt and Christian Werner. 2007. Network music performance-problems, approaches and perspectives. In Proceedings of the “Music in the Global Village”-Conference, Budapest, Hungary, Vol. 162. 23–10.
  • Durner et al. ([n. d.]) Raphael Durner et al. [n. d.]. Performance study of dynamic QoS management for OpenFlow-enabled SDN switches. In IEEE IWQoS’15.
  • Durner et al. (2015) Raphael Durner, Andreas Blenk, and Wolfgang Kellerer. 2015. Performance study of dynamic QoS management for OpenFlow-enabled SDN switches. IEEE, 177–182.
  • Egilmez and Tekalp (2014) Hilmi E. Egilmez and A. Murat Tekalp. 2014. Distributed QoS Architectures for Multimedia Streaming Over Software Defined Networks. IEEE Transactions on Multimedia 16, 6 (Oct. 2014), 1597–1609.
  • Gkounis et al. (2016) D. Gkounis, V. Kotronis, C. Liaskos, and X. Dimitropoulos. 2016. On the Interplay of Link-Flooding Attacks and Traffic Engineering. ACM SIGCOMM Computer Communication Review 46, 1 (2016), 5–11.
  • Gorlatch et al. ([n. d.]) Sergei Gorlatch et al. [n. d.]. Improving QoS in real-time internet applications. In IEEE ICCNC’14.
  • Goto et al. (1910) Masataka Goto, Isao Hidaka, Hideaki Matsumoto, Yosuke Kuroda, and Yoichi Muraoka. 1910. A virtual jazz session system: VirJa session. Transactions of Information Processing Society of Japan 21 (1910).
  • Kaur et al. (2014) Sukhveer Kaur, Japinder Singh, and Navtej Singh Ghumman. 2014. Network programmability using POX controller. In ICCCS International Conference on Communication, Computing & Systems, IEEE. 138.
  • Kobayashi et al. (2014) Masayoshi Kobayashi, Srini Seetharaman, Guru Parulkar, Guido Appenzeller, Joseph Little, Johan van Reijendam, Paul Weissmann, and Nick McKeown. 2014. Maturing of OpenFlow and Software-defined Networking through deployments. Computer Networks 61 (March 2014), 151–175.
  • Kotronis et al. (2014) Vasileios Kotronis, Xenofontas Dimitropoulos, et al. 2014. Control exchange points: Providing qos-enabled end-to-end services via sdn-based inter-domain routing orchestration. LINX 2429, 1093 (2014), 2443.
  • Kotronis et al. (2016) Vasileios Kotronis, Rowan Klöti, Matthias Rost, Panagiotis Georgopoulos, Bernhard Ager, Stefan Schmid, and Xenofontas Dimitropoulos. 2016. Stitching inter-domain paths over IXPs. In Proceedings of the Symposium on SDN Research. ACM, 17.
  • Koumaras et al. (2016) Harilaos Koumaras, Michail-Alexandros Kourtis, Christos Sakkas, George Xilouris, and Stavros Kolometsos. 2016. In-service Video Quality assessment based on SDN/NFV techniques. IEEE, 1–5.
  • Kumar et al. (2013) Himal Kumar, Hassan Habibi Gharakheili, and Vijay Sivaraman. 2013. User control of quality of experience in home networks using SDN. IEEE, 1–6.
  • Lantz et al. ([n. d.]) Bob Lantz et al. [n. d.]. A network in a laptop. In HotNets-IX (2010).
  • Lazzaro and Wawrzynek ([n. d.]) John Lazzaro and John Wawrzynek. [n. d.]. A case for network musical performance. In ACM NOSSDAV’01.
  • Liaskos (2015) C. Liaskos. 2015. A lightweight, non-intrusive approach for orchestrating autonomously-managed network elements. In IEEE ISCC’15. IEEE, 335–340.
  • Liaskos et al. (2016a) C. Liaskos, X. Dimitropoulos, and L. Tassiulas. 2016a. Backpressure on the Backbone: A Lightweight, Non-intrusive Traffic Engineering Approach. IEEE Transactions on Network and Service Management to appear (2016), 1–14.
  • Liaskos et al. (2016b) C. Liaskos, V. Kotronis, and X. Dimitropoulos. 2016b. A Novel Framework for Modeling and Mitigating Distributed Link Flooding Attacks. In IEEE INFOCOM’16. 1–9.
  • Maribondo and Fernandes ([n. d.]) P. Maribondo and N. Fernandes. [n. d.]. Avoiding Voice Traffic Degradation in IP Enterprise Networks Using CAoS. In ACM LANCOMM’16.
  • McKeown et al. (2008) Nick McKeown, Tom Anderson, Hari Balakrishnan, Guru Parulkar, Larry Peterson, Jennifer Rexford, Scott Shenker, and Jonathan Turner. 2008. OpenFlow: enabling innovation in campus networks. ACM SIGCOMM Computer Communication Review 38, 2 (March 2008), 69.
  • Mu et al. (2016) Mu Mu, Matthew Broadbent, Arsham Farshad, Nicholas Hart, David Hutchison, Qiang Ni, and Nicholas Race. 2016. A scalable user fairness model for adaptive video streaming over SDN-assisted future networks. IEEE Journal on Selected Areas in Communications 34, 8 (2016), 2168–2184.
  • Nam et al. (2014) Hyunwoo Nam, Kyung-Hwa Kim, Jong Yul Kim, and Henning Schulzrinne. 2014. Towards QoE-aware video streaming using SDN. IEEE, 1317–1322.
  • Nurmela (2007) Tuomas Nurmela. 2007. Session Initiation Protocol. In Seminar on Transport of multimedia streams, University of Helsinki. Citeseer.
  • Prete et al. ([n. d.]) L. R. Prete et al. [n. d.]. Simulation in an SDN network scenario using the POX Controller. In IEEE COLCOM’14.
  • Schuett (2002) Nathan Schuett. 2002. The effects of latency on ensemble performance. Bachelor Thesis, CCRMA Department of Music, Stanford University (2002).
  • Sharma and Sood (2014) Kuldeep K Sharma and Manu Sood. 2014. Mininet as a Container Based Emulator for Software Defined Networks. International Journal of Advanced Research in Computer Science and Software Engineering 4, 12 (2014).
  • Sharma et al. (2014a) Sachin Sharma, Dimitri Staessens, Didier Colle, David Palma, Joao Goncalves, Ricardo Figueiredo, Donal Morris, Mario Pickavet, and Piet Demeester. 2014a. Implementing Quality of Service for the Software Defined Networking Enabled Future Internet. IEEE, 49–54.
  • Sharma et al. (2014b) Sachin Sharma, Dimitri Staessens, Didier Colle, David Palma, Joao Goncalves, Mario Pickavet, Luis Cordeiro, and Piet Demeester. 2014b. Demonstrating resilient quality of service in Software Defined Networking. IEEE, 133–134.
  • Sieber et al. (2015) Christian Sieber, Andreas Blenk, David Hock, Marc Scheib, Thomas Hohn, Stefan Kohler, and Wolfgang Kellerer. 2015. Network configuration with quality of service abstractions for SDN and legacy networks. IEEE, 1135–1136.
  • Silvan and Christos ([n. d.]) Streit Silvan and Kalialakis Christos. [n. d.]. A POX OpenFlow Loop Solution for Mininet Network Emulations.
  • Sinnreich and Johnston (2006) Henry Sinnreich and Alan B. Johnston. 2006. Internet communications using SIP: delivering VoIP and multimedia services with Session Initiation Protocol (2nd ed ed.). Wiley Pub, Indianapolis, IN. OCLC: ocm65340953.
  • Sukhveer et al. ([n. d.]) Kaur Sukhveer, Singh Japinder, and Singh Ghumman Navtej. [n. d.]. Network Programmability Using POX Controller.
  • Tomovic et al. (2014) Slavica Tomovic, Neeli Prasad, and Igor Radusinovic. 2014. SDN control framework for QoS provisioning. IEEE, 111–114.
  • Valin et al. (2016) Jean-Marc Valin, Gregory Maxwell, Timothy B Terriberry, and Koen Vos. 2016. High-quality, low-delay music coding in the opus codec. arXiv preprint arXiv:1602.04845 (2016).
  • Wabnik et al. (2009) Stefan Wabnik, Gerald Schuller, and Ferenc Kraemer. 2009. An error robust ultra low delay audio coder using an MA prediction model. In Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE International Conference on. IEEE, 5–8.
  • Xiaoyuan Gu et al. (2005) Xiaoyuan Gu, M. Dick, Z. Kurtisi, U. Noyer, and L. Wolf. 2005. Network-centric music performance: practice and experiments. IEEE Communications Magazine 43, 6 (June 2005), 86–93.