Institute of Computer Science Chair of Communication Networks Prof. Dr.-Ing. P. Tran-Gia Dynamic Bandwidth Allocation for Multiple Network Connections: Improving User QoE and Network Usage of YouTube in Mobile Broadband Florian Wamser, Thomas Zinner, Phuoc Tran-Gia University of Würzburg, Germany Jing Zhu Intel Labs, Hillsboro, OR, United States
Competing Applications at a Bottleneck Link Voice over IP Online Office YouTube Video Software Download Internet % 2 Florian Wamser 2
Impact on Application Quality Content unaware networks � � Fair share with respect to QoS (throughput) Bulk data download performance: good � YouTube quality: bad � 400 15 Throughput Desired 300 ] progress progress s buffered playtime [s] / bulk data YouTube k B 10 (downloads) Influence on [ 200 YouTube Actual Actual a t progress progress d a 5 100 0 0 0 10 20 30 40 50 60 0 20 40 60 time [s] time [s] % 3 Florian Wamser 3
Application-Aware Networking � Tasks and objectives Application and � Integrating application needs’ in network resource management Network � Add or re-allocate resources on demand Monitoring Application and network monitoring 1. � Collects information with high correlation to QoE � Example: YouTube monitor (YoMo), browsing monitor, etc. Decision Entity Decision entity 2. � Evaluates the information and decides about appropriate resource management action Dynamic resource management 3. � Enforces resource management actions Dynamic � Example: resource allocation, scheduling, traffic prioritization, Resource access technology selection, … Management % 4 Florian Wamser 4
Resource Management: Dual Connectivity of Devices � More than just one transmission technology is available at current mobile devices � Wi-Fi Communications � Cellular Communications % 5 Florian Wamser 5
Framework for Intelligent Bandwidth Aggregation � Virtual access network (VAN) to aggregate multiple networks into single IP pipe � Technical implementation: TCP/IP over UDP tunneling (mobile IP-like approach) � Features of Intel‘s OTT VAN � Configurable bandwidth aggregation for multiple networks � (TCP) packet reordering (re-sequencing) � Missing features � Smart algorithms for dynamic offloading � Application specific guidelines % 6 Florian Wamser 6
Intel’s OTT VAN Testbed: Hardware and Software Client Server � Virtual network device provides tunneling � Tunnel endpoint functionality � Implements re-sequencing � Access technologies buffer � Wi-Fi communications (limited to max. 2 Mbps) � Enforces resource management � Mobile communications (limited to 4 Mbps) Mobile Access Network Application TCP/UDP Core IP Network UDP UDP IP IP Internet Application Wi-Fi (Wi-Fi) (3G) Access TCP/UDP Network IP % 7 Florian Wamser 7
Resource Management Algorithms � Adjust offload ratio between Wi-Fi and 3G cellular traffic, based on a required throughput � Always use Wi-Fi and dynamically add 3G If current throughput < required throughput � Increase 3G bandwidth If current throughput > required throughput � Decrease 3G bandwidth % 8 Florian Wamser 8
Resource Management Algorithms for � Algorithm 1: Static Offloading Based on Video Request � Defines required throughput based on requested Algorithm 1 video quality Req. throughput Detects uplink request by YouTube with DPI � Time � Algorithm 2: Dynamic Offloading Based on Buffer Estimation Algorithm 2 � Constant monitoring of the buffer level Req. throughput � Adaption of the required throughput based on the buffer level Time req. throughput = max. � Algorithm 3: Burst-wise Offloading Based on Buffer Estimation � Make use of the complete bandwidth until the Algorithm 3 buffer is filled � Disables 3G link until the buffer gets low Time req. throughput = 0 � WiFi-only % 9 Florian Wamser 9
Time Series of Algorithm 1 � Time series of one video with 1080p resolution � Wi-Fi and 3G available � Static req. throughput = 6 Mbps 50 Buffered time [s] 40 30 Buffer filled 20 10 Request detected 0 0 50 100 150 200 250 300 Time [s] Throughput [Mbps] 6 5 4 Adaption to No WiFi or 3G activity required throughput 3 2 Request detected 1 0 0 50 100 150 200 250 300 Time [s] % 10 Florian Wamser 10
Time Series of Algorithm 2 � Algorithm 2 dynamically adjusts required throughput according to video playback buffer 50 Buffered time [s] 40 30 Buffer filled 20 10 Request detected Low buffer 0 0 50 100 150 200 250 300 Time [s] Throughput [Mbps] 6 5 4 3 2 1 Request detected Throughput decreases 0 0 50 100 150 200 250 300 Time [s] % 11 Florian Wamser 11
Comparison of Algorithm 1 and 2 � Consumption of 3G bandwidth Algorithm 1 (resolution based) Algorithm 2 (buffer based) � per 1000 seconds playtime Total 3G bandwidth consumed in MB Total 3G bandwidth consumed in MB 400 400 1080p 720p 300 300 200 200 100 100 0 2.5 3 3.5 4 4.5 5 1.6 1.8 2 2.2 2.4 2.6 Average bit-rate in Mbps Average bit-rate in Mbps Resolution based Buffer based 720p 98 MB -64,3% 35 MB 321 MB 223 MB 1080p -30,5% % 12 Florian Wamser 12
Comparison of Algorithm 1 and 2 � Average amount of consumed energy J. Huang et al. „A close examination of performance and power characteristics of 4g lte networks” Energy consumption per video in Joule � per 1000 seconds playtime Energy consumption per video in Joule 1600 1400 720p 1080p 1400 1200 1200 1000 1000 800 800 600 600 Resolution based 400 400 Buffer based 200 200 2.5 3 3.5 4 4.5 5 1.6 1.8 2 2.2 2.4 2.6 Average bit-rate in Mbps Average bit-rate in Mbps Resolution based Buffer based -49,9% 1180 J 591 J 720p 1404 J 1299 J 1080p -7,5% % 13 Florian Wamser 13
Time Series of Algorithm 3 � Algorithm 3 activates 3G link in bursts 50 Full Buffer Buffered time [s] Full Buffer 40 30 20 Request 10 detected Low Buffer Low Buffer 0 0 50 100 150 200 250 300 Time [s] Throughput [Mbps] 6 5 4 3 Request detected 2 1 3G link disabled 0 0 50 100 150 200 250 300 Time [s] % 14 Florian Wamser 14
Comparison of Algorithm 2 and 3 � Average amount of consumed energy Dynamic buffer Energy consumption per video in Joule Energy consumption per video in Joule Burst-wise buffer � per 1000 seconds playtime 1600 1400 720p 1080p 1400 1200 1200 1000 1000 800 800 600 600 400 400 200 200 2.5 3 3.5 4 4.5 5 1.6 1.8 2 2.2 2.4 2.6 Average bit-rate in Mbps Average bit-rate in Mbps Dynamic buffer Burst-wise buffer -12,2% 591 J 519 J 720p 1299 J -8,9% 1183 J 1080p % 15 Florian Wamser 15
Conclusion and Outlook � Contribution of the work � Assessment and quantification of the benefits of cross-layer resource management on the example of YouTube � Analysis of three application-aware algorithms which differ in complexity and impact on user and network � Results of the evaluation � The application-aware algorithms can – enhance the QoE level for end users (if both networks provide enough resources) – save costs in terms of energy & Cellular resources � Future work � Investigations on scalability of our approach and field trials with many users � Providing a holistic resource allocation for popular applications with respect to their instaneneous needs % 16 Florian Wamser 16
http://dl.acm.org/authorize?N71341 Florian Wamser, Thomas Zinner, Phuoc Tran-Gia and Jing Zhu Dynamic Bandwidth Allocation for Multiple Network Connections: Improving User QoE and Network Usage of YouTube in Mobile Broadband % 17 Florian Wamser 17
http://dl.acm.org/authorize?N71209 Florian Wamser, Thomas Zinner, Lukas Iffländer, Phuoc Tran-Gia Demonstrating the Prospects of Dynamic Application-Aware Networking in a Home Environment % 18 Florian Wamser 18
http://dl.acm.org/authorize?N71235 Steffen Gebert, David Hock, Thomas Zinner, Phuoc Tran-Gia (University of Würzburg); Marco Hoffmann, Michael Jarschel, Ernst-Dieter Schmidt (Nokia); Ralf-Peter Braun (Deutsche Telekom T-Labs), Christian Banse (Fraunhofer AISEC); Andreas Kopsel (BISDN) Demonstrating the Optimal Placement of Virtualized Cellular Network Functions in Case of Large Crowd Events % 19 Florian Wamser 19
Recommend
More recommend
