The Networking Lab in the School of Computing 1 Dr. Jim Martin Associate Professor School of Computing Clemson University Networking Lab’s Website: Don’t worry Homer, the guys/girls in the Networking Lab will fix things!! WooHoo….I don’t know what this means There are no more /8 TCP/IP V4 Addresses!!

Research Group’s Mission Vision Statement: – Computing and the Internet are converging – Traditional broadcast video (Cable/Satellite) is converging with the Internet – Networks are becoming more and more ‘heterogeneous’ – The scope of the lab’s interests is more than networking, it includes operating systems, distributed systems, secure and trustworthy systems, and next generation Internet. Collectively these define the term ‘cybersystems’ The networking group focuses on a range of problems that are at the heart of developing and analyzing emerging cybersystems. The mission of the lab is to support cutting edge research in cybersystems AND to train researchers to address the needs of the changing world.. 2 Now that’s what I’m talking about!

Brief Introduction The Networking Group focuses on a range of problems in the area of computer networking. The focus the last several years has been on broadband Internet access. – Web site: Funding – NSF, Cisco, CableLabs, Department of Justice/National Institute of Justice, NASA, IBM Team – Faculty: Jim Martin, Mike Westall, Brian Dean, Juan Gilbert, KC Wang, Harlan Russel, Richard Brooks – Students: Rahul Amin, Yunhui Fu, Gongbing Hong, and many MS and undergrad students 3

NETLAB Activities 4 ResearchPedagogyOutreach Wireless Systems Broadband Access Internet Protocols and Issues Networking and Systems Course Development VM-based labs Emerging Scholars CyberTiger Creative Inquiry Statewide Broadband Wireless Initiatives Networking Seminar Broadband Service Website Testbeds and Experimental Deployments

5 CyberInfrastructure CyberSystems Research Application Domain (e.g., connected vehicle) Trustworthy computing System Performance Computational Theory AlgorithmsSoftware Engineering CyberInfrastructure: The hardware/software systems that operate harmoniously to meet the requirements of domain specific applications and systems. CI includes operating systems, networks, distributed systems, secure and trustworthy systems, HPC.

Internet Wired Network Video Entertainment (YouTube, Netflix Hulu, …) WiFi Web, Cloud applications 3G/4G Cellular Data Service (e.g., Verizon or AT&T) Wired Access (e.g., Cable, DSL, Fiber) Smart Home Network Power Company’s Smart Grid Support Sensor Net Satellite Access This shaded cloud is where broadband service providers such as Comcast and AT&T would like to evolve towards (a true service provider not just a network provider) Wireless Heterogeneous Networks (aka hetnets) extend application domains over a wireless Internet – Driven by emerging application domains such as Intelligent Transportation System, Environmental Sensing, SmartGrid, health care, mobile broadband wireless, …. Challenge #1 : Cooperative networks are COMPLICATED Challenge #2 : Cooperative networks BREAK current ISP economic models Challenge #3 : The current method for spectrum management is OUTDATED Challenge #4 : Pushes the original TCP/IP design BEYOND its capabilities. Current home environments are not too far from this today!! The science behind this work is in the area of resource allocation and systems performance

CyberTiger Creative Inquiry (formerly iTiger) : Engage undergrads to work with PhD students on projects that develop applications that facilitate the study of large scale broadband wireless systems Current focus is building tools (server back end, smartphone apps) that provide a measurement and assessment capability of broadband cellular and WiFi data services Recent activities include a report to the FCC on the LACK of cellular infrastructure in SC outside of populated areas The contribution of this work is the broader impact on society- we want to ensure everyone in the USA has equal access to measurably robust wireless services

Netlab’s Wireless Research 8 – Building Cooperative Heterogeneous Wireless Networks With Re-Configurable Devices (NSF) The federal government is investing lots of money in how open spectrum can be used to enhance and extend broadband wireless access. We are developing techniques by which future multi-radio, reconfigurable smart phone devices can use multiple blocks of spectrum at any given time. – We would like to develop an experimental solution to the WiFi offloading problem. – Broadcasting Video Content in Dense g Sports and Entertainment Venues (Cisco) Cisco Sports and Entertainment has a solution where video camera feeds in pro-stadium venues are broadcast over g in-stadium network. We are developing an adaptive application FEC technique that maximizes APFEC effectiveness while minimizing overhead. – How might this work over Verizon’s 4G network? – We would like to develop an experimental solution to this WiFi offloading problem.

Netlab’s Wireless Research 9 – CyberTiger Creative Inquiry Undergrads are developing a broadband wireless mapping capability to assess the coverage and performance of 3G/4G in SC. – We would like to extend the scope of the study to monitor student usage of broadband wireless on campus. – Perceived Quality User Assessment We are conducting human-computer interface studies that will provide guidelines for acceptable range of application behaviors over broadband access links. Applications include networked games and TCP-based adaptive bitrate video streaming. – We would like to extend the scope of the study to assess how types of users perceive the utility of smartphones subject to different device form factors, user interfaces, service quality levels.

Building Cooperative Heterogeneous Wireless Networks With Re-Configurable Devices 10 Internet Access Network Exit SmartPhone Global Resource Controller Autonomous Wireless Systems Open Spectrum Future handhelds will contain multiple radios that can be used concurrently AND that are reconfigurable. Future wireless networks will be heterogeneous with cooperative mechanisms in place (early examples are femtocells and WiFi off loading)

Problem Statement: Future Heterogeneous Wireless Systems 11 Carrier 2: WiFi Carrier 1: WiFi Carrier 2: WiFi Carrier 1: WiMAX Carrier 2: LTE Carrier 2: HSPA Carrier 1: EVDO

Dynamic Adaptive Streaming over HTTP (DASH) 12 Let’s rewind a bit: Video streaming has evolved from UDP-based streaming, to TCP-based progressive downloading, to adaptive HTTP- based streaming Several well established methods for adaptive HTTP streaming: Adobe’s HTTP Dynamic Streaming (HDS) Apple’s HTTP Live Streaming (HLS) Microsoft’s Smooth Streaming (SmoothHD) Will Law, Akamai's principal architect for media engineering, commented, “We've spent the past five years delivering a variety of adaptive video formats—SmoothHD, HLS and HDS—all of which are 80 percent the same but 100 percent incompatible.”

DASH 13 MPEG organization issued a Call for Proposal for an HTTP Streaming Standard (2009). Resulting standard is MPEG-DASH over HTTP and has been published as standard ISO/IEC The 3GPP community has been involved since GP-DASH is in 3GPP standards TS and 3GPP TS The initiative has significant industry support…. although not exactly clear about Apple or YouTube. Let’s fast forward a bit…. DASH will make it easier for a content provider to build large scale distribution systems that support the majority of client device types. DASH is encoder/decoder agnostic. Content providers will have to select the video encoding options and the range of supported bit rates.

MPEG DASH 14 Figure from presentation at Streaming Media West “MPEG-DASH: Driving the Growth of Streaming Using the New HTTP Standard (available online at )

MPEG DASH 15 Figure from presentation at Streaming Media West “MPEG-DASH: Driving the Growth of Streaming Using the New HTTP Standard (available online at )

MPEG DASH 16 A segment is an independent, viewable period of video/audio/timing data. Segment size of 2 seconds or 10 seconds is reasonable. Segments are uniquely identified by an HTTP URL. A client requests the segment, the bit rate, and optionally a specific byte range in the segment. Clients can issue requests and receive segments over any number of concurrent TCP connections. The video segment is sent back by the HTTP server in a ‘burst’. The implementation of the client determines how frequently segments are requested, when bit rate adaptation occurs, and the overall ‘sensitivity’ of the application to network congestion.

17 Buffering state (37.02 Mbps between seconds) Steady state (4.39 Mbps) Background: DASH Behavior Looking at the first 80 seconds of a Netflix session on a Windows Desktop device. There are multiple ‘startup’ TCP/HTTP connections that establish the session (Netflix servers) Once the session is established, a CDN server is selected, and this client initiates a single TCP connection to transfer ALL content (video and audio) Trace captured at the client network (all TCP flows captured) We see a high initial throughput as the client fills the playback buffer This is followed by a steady state throughput of 4.39 Mbps Upstream throughput during the 80 seconds was about 1 Kbps. The ratio of downstream traffic to upstream traffic was 3336/1. The ratio of packets sent downstream to upstream was 2.68 a.TCP Throughput of single Cx used for content (Windows device, high speed Internet access (>100Mbps), very good network conditions)

18 Buffering state Steady state a. TCP Sequence / Ack Plot Time Period Highlighted in Figure b. Background: DASH Behavior Black shows packet arrivals (Red shows ACK’s – but hard to see) During steady state, client and server exhibit on/off behavior. After time 30 seconds, mean (standard deviation) On period bytes transferred: 2,217,200 bytes ( bytes) On period peak throughput: 4.49 Mbps (0.50 Mbps) On/Off duty cycle time : seconds ( seconds) We conjecture the Netflix video encoder is set to CBR with some variability allowed. Cycle Time : 4.36 seconds (mean during steady state) On Period Burst Length: 2.14 Mbytes (mean during steady state) b. Time Seconds Enlarged

Objectives 19 We wanted to characterize the bandwidth consumption and behaviors of an early DASH implementation : Netflix Are there differences across a range of client devices? Open questions are: How does application level control co-exist with TCP control? Does it give up too much bandwidth? Is it well behaved and stable during periods of volatile network conditions? What are the main design parameters at the client that impact performance ?

Methodology 20 Two components to the study Measurement study - conduct controlled experiments using live Netflix sessions. Analyzed tcpdump packet captures Used 4 different client devices Simulation study - developed a simulation model of a DASH application using the ns2 network simulator Dig deeper to better understand empirical results Explore the design space of the client Primary performance metrics: Aggregate bandwidth consumed by the session under observation Throughput and behavior of individual TCP connections

Methodology : Measurement Testbed

Go Back to the Basics 22 Resource Allocation: The process by which network elements try to meet the competing demands that applications have for network resources

Wrap Up…..Final Message – The Internet is evolving because of technology but also because of economic and societal change – Tremendous changes are happening due to Convergence of cable/over-the-air broadcasting, the Internet, and Mobile devices Convergence of applications, operating systems, and the Internet – Human Centered Computing is very evident in the Internet…. As networks connect humans, the effects of HCC are social interactions, social-systems effects, and the subsequent economics that attempt to capitalize on how things play out. – Final thoughts: Our broad focus is to develop a theoretical framework for building networks of the future that meet new requirements Preparation for research in networking: required: CPSC851,852,854, helpful: math classes on optimization, random processes, ECE 848. Please contact me if you would like to discuss our groups direction 23