1 Applying Adaptive Traffic Management: White Paper Applying Adaptive Traffic Management: Improving Network Capacity and the Subscriber Experience citrix.com
2 i Applying Adaptive Traffic Management: Improving Network Capacity and the Subscriber Experience Table of Contents Executive Summary... 3 Adaptive Traffic Management... 3 Adaptive Traffic Management Improves the Network for Subscribers and Operators... 4 Step One: Detect Step Two: Decide... 6 Step Three: React... 6 Adaptive Traffic Management versus Deep Packet Inspection... 6 Adaptive Traffic Management Is Subscriber-Centric... 6 Adaptive Traffic Management Implementation Scenarios... 8 Adaptive Traffic Management Based on Subscriber QoE... 8 Adaptive Traffic Management Based on RAN Input... 9 Static Congestion Scenario Dynamic Congestion Scenario Adaptive Traffic Management Based on a Specific Application Adaptive Traffic Management Based on a Global View of Subscribers Deploying Adaptive Traffic Management G Networks LTE/4G Networks Conclusion Abbreviations Appendix: Underlying Techniques of Adaptive Traffic Management Application Detection and Control Content-Layer Control Citrix ByteMobile Content-Layer Optimization Content Caching WP-ATM-0822-A
3 3 Executive Summary Adaptive Traffic Management dynamically applies the appropriate traffic management techniques, at the right time, to the subscribers who need it As smart mobile devices become ubiquitous, mobile network operators (MNOs) are at a turning point. Networks are becoming stressed by the exponential growth in demand for mobile bandwidth growth that is both unpredictable and transient. The resulting poor page download speeds and stalling videos mean disappointed and frustrated subscribers. Because mobile networks must operate within defined radio interface and spectrum restrictions, mobile network operators are looking at all areas of potential efficiency gains within the network managing network traffic is a critical requirement. Adaptive Traffic Management dynamically applies the appropriate traffic management techniques, at the right time, to the subscribers who need it based on policy inputs, subscriber quality of experience (QoE) requirements and the condition of the radio access network (RAN). This allows mobile network operators (MNOs) to: deliver the highest possible QoE to all subscribers extend the capacity of existing 3G networks control the impact of over-the-top (OTT) video traffic in new 3rd Generation Partnership Project (3GPP) long-term evolution (LTE) or 4G (LTE-Advanced) networks increase revenue Adaptive Traffic Management Citrix ByteMobile s Adaptive Traffic Management offers MNOs a solution to the challenge created by the explosion in mobile network traffic. Leveraging holistic information about each subscriber s application, web and video traffic, Adaptive Traffic Management applies sophisticated traffic management techniques to help MNOs deliver the highest possible QoE to all subscribers. By dynamically monitoring traffic for each subscriber from the IP layer to the content layer Adaptive Traffic Management intelligently adjusts subscribers traffic based on policy inputs, applications used by each subscriber, real-time subscriber QoE, and the changing condition of the radio access network (RAN). Adaptive Traffic Management monitors RAN conditions and tracks the user experience index (UXI) of each subscriber by measuring experience-related variables such as video stalling, video resolution What is QoE? Quality of experience is a measure of a subscriber s perception of and satisfaction with a service for example, web browsing, file downloads or video playback. QoE is related to, but differs from, quality of service (QoS), which is a measure of the service delivered at the network level. QoE can be thought of as perceived QoS: the quality of the subscriber s experience is based only on what the subscriber perceives, not the network-level measurement of the service. For example, does a subscriber experience stalling, blockiness, blur, or jerkiness in a video, regardless of the download speed? Mobile video content is fast becoming the majority of all mobile network traffic, so a subscriber s QoE when watching video is of utmost importance. QoE is the single-most significant factor in a realworld evaluation of the subscriber experience. Subscribers now expect a high-quality, TV-like viewing experience a video connection that is smooth, without any annoying stalling whenever they want it, on whatever device they choose, wherever they are.
4 4 and web object download speed all in real time. To deliver the highest possible QoE to all subscribers, one or more of the following traffic management techniques is selectively and dynamically applied: Application detection and control: Identifying applications used by each subscriber and applying flow and session-based blocking, shaping and marking Content layer control: Applying application and automated-software-download control, just-in-time (JIT) video delivery, streaming media policy control and lossless web optimization Content-layer optimization: Applying Quality-Aware Transcoding (QAT) with standards-based video compression techniques and Dynamic Bandwidth Shaping (DBS) Content caching: Storing video, web and software auto-update content closer to the subscriber Dependent on the individual subscriber experience, traffic management techniques are dynamically selected and applied. A unique approach to traffic control, Adaptive Traffic Management combines network intelligence and awareness of the application and content layers (L4 L7) with the ability to take immediate corrective action. With the power to detect and respond to network anomalies before they affect subscribers, MNOs using Adaptive Traffic Management can offer a consistent, high-quality experience for all subscribers. Adaptive Traffic Management Improves the Network for Subscribers and Operators Adaptive Traffic Management intelligently uses traffic management techniques to maximize network resources and improve the subscriber experience. Adaptive Traffic Management helps improve subscriber QoE and deliver a more deterministic mobile data service by applying the right traffic control techniques, in the right
5 5 amount, to the subscribers who need it, at the right time. Adaptive Traffic Management applies appropriate traffic management techniques in a targeted fashion based on network policy inputs, subscriber QoE and RAN conditions, in these three steps: Detect: Collect and monitor subscriber experience and application traffic, as well as network conditions, in real time Decide: Determine the correct level of traffic management to optimize QoE for all subscribers, based on current subscriber QoE and network conditions React: Intelligently apply the appropriate traffic management technique at the right time In this way, Adaptive Traffic Management helps MNOs dynamically address poor network conditions and the impact of multiple subscribers in a cell to maintain the best possible QoE for all subscribers. Adaptive Traffic Management is applied in 3 steps: Detect, Decide and React. Step One: Detect To ensure a high QoE for all subscribers, Adaptive Traffic Management uses a single platform to collect and monitor subscriber and application traffic in real time, including: User experience index for each subscriber: Monitoring factors such as video stalling, video resolution, web page download time, and file download time to determine the user experience index (UXI), a numerical expression of the quality of experience for each subscriber that can change over time. The UXI represents holistic knowledge of the user experience measured dynamically for all subscribers on a cell and allows for decisions that help ensure efficient control of resources under network load. External network Intelligence: Leveraging RAN intelligence data directly from RAN elements, network probes or other sources that may indicate a congestion condition and the subscribers who are affected.
6 6 Global view of applications: Identifying applications used by each subscriber; for example, , P2P, web traffic, messaging, streaming video or audio, progressive download video, social media or gaming. Step Two: Decide Adaptive Traffic Management decides what action to take based on the UXI and a global view of applications traversing network hotspots, selectively applying traffic management techniques web optimization, application detection and control, content layer control and content layer optimization to prevent stalling, create headroom and ensure the highest possible QoE for all subscribers. Step Three: React Finally, Adaptive Traffic Management reacts. Essentially, Adaptive Traffic Management intelligently applies appropriate traffic management techniques, starting with the least intensive, and graduating to the most intensive when network conditions indicate the need. Adaptive Traffic Management versus Deep Packet Inspection A Deep Packet Inspection (DPI) tool typically processes only a few packets of each subscriber application flow, using application detection capabilities to identify and classify the application. The DPI tool then hands off the IP flow to a network processor to apply shaping, blocking, or packet marking, if necessary. Once the DPI tool has classified the flow, it essentially loses sight of the flow, as it does not examine any subsequent packets. This becomes a problem in mobile networks, where dynamically changing network conditions are the norm. In addition, DPI tools cannot consistently identify complex embedded traffic like progressive download videos. To do so, the DPI tool would need to hold packets longer, significantly reducing the amount of traffic the DPI tool could handle. In the limited number of cases in which a DPI tool can identify the flow, the DPI tool only applies bulk traffic management functions to the entire flow. For flows like video, whose bit rate varies based on the content being delivered for example, high-motion versus still scenes DPI tools do not control each packet and cannot preserve high QoE. Adaptive Traffic Management Is Subscriber-Centric DPI tools are network-centric, and, as such, DPI tools may specify generic corrective actions that do not rectify the situation, can waste network resources, and can negatively impact other subscribers, potentially degrading their service. For example, if a RAN is congested, some DPI tools cap service delivery for an entire application class, thereby negatively impacting subscribers who are not contributing to the issue. Some DPI tools can limit service across all traffic from certain
7 7 subscribers, including applications that are not creating problems, thereby impairing the performance of the non-problematic applications. Unlike DPI tools, Adaptive Traffic Management is fully aware of the subscriber, application and content. Adaptive Traffic Management processes each packet of every session and does not lose sight of the session after classifying it. Adaptive Traffic Management dynamically applies various traffic management techniques throughout the subscriber session, depending on the nature of the session, the dynamic condition of the underlying network and the resultant QoE being delivered. Adaptive Traffic Management is subscriber QoE centric, enabling MNOs to gain a better understanding of the specific interactions between IP traffic and RAN congestion and to better satisfy subscribers, reduce churn and increase revenue. Adaptive Traffic Management is Subscriber-Experience Centric.
8 8 Adaptive Traffic Management Implementation Scenarios Adaptive Traffic Management offers several solutions to ensure that MNOs can consistently deliver a high QoE to all subscribers. Adaptive Traffic Management can be based on: Subscriber QoE RAN input A specific application A global view of subscribers Adaptive Traffic Management Based on Subscriber QoE Because QoE the subjective measure of a subscriber s experiences with a service is of prime importance to both subscribers and MNOs QoE is a major consideration for Adaptive Traffic Management. Envision a scenario in which a subscriber is watching a video and the traffic on the subscriber s cell increases significantly, perhaps because many other subscribers sign on or some existing subscribers initiate activities that consume large amounts of bandwidth. The additional network traffic generated by the other subscribers sessions negatively impacts the first subscriber s video connection, stalls the video and thus generates a poor UXI. In this scenario, Adaptive Traffic Management will employ content optimization essentially, serving a video with increased video compression if a subscriber requests another video within a specified period of time thereby reducing the load on the cell. Adaptive Traffic Management based on subscriber QoE.
9 9 Table 1. Adaptive Traffic Management Based on Subscriber QoE Scenario Condition Detect Decide React Subscriber is watching a video. Traffic on the subscriber s cell increases significantly due to additional traffic from other subscribers. If no prior history exists, track UXI; if prior history exists, update UXI. Detect the onset of video stalling. If UXI is not satisfactory, decide to apply content optimization. As the subscriber continues further video sessions, Adaptive Traffic Management employs content optimization to prevent stalling. Adaptive Traffic Management Based on RAN Input At the network core, traffic is aggregated, which has a slight smoothing effect and makes peak loading more predictable. However, network traffic at individual cell sites is much less predictable. At the cell level, traffic volumes, applications and subscribers vary greatly from site to site, from day to day, and by time of day, making congestion a local, transient phenomenon. The reality of mobile broadband networks is that bandwidth usage is constantly changing a certain cell gets unpredictably congested one minute and is at only a small percentage of capacity the next. However, in many cases MNOs know when certain cells will experience heavy loads for example, the cells around a stadium during a playoff game or the cells around a major shopping venue during a major holiday weekend. This combination of predictable and unpredictable congestion leads to two Adaptive Traffic Management scenarios based on RAN input: Static: MNOs predict congestion of a specific cell by day and time Dynamic: Random congestion of a cell is indicated by a RAN element or network probe Adaptive Traffic Management based on RAN imput.
10 10 Static Congestion Scenario An MNO may know that a specific cell will be congested on a specific day at a specific time, based on historical projections or previous knowledge of a specific event. In this case, based on that projection, the MNO can designate a location ID and time period as congested. Adaptive Traffic Management can use these projections to proactively apply traffic management techniques to all subscriber sessions for the specified location and time period, thereby protecting subscribers QoE before they experience any problems. Picture a scenario in which an MNO knows many people will congregate in a specific location for example, the World Cup final at the Maracanã Stadium in Rio de Janeiro, Brazil on July 13, 2014 at 16:00. The MNO knows that many subscribers gathered in one place at one time are certain to generate a great deal of network traffic. In this scenario, Adaptive Traffic Management proactively applies traffic management for all subscribers, reducing the load on the cell and ensuring the highest possible QoE for all subscribers. Table 2. Adaptive Traffic Management Based on Static RAN Input Scenario Condition Detect Decide React MNO knows that many subscribers will assemble in a specific place at a specific time. MNO specifies the location and time of predicted network congestion. Detect the arrival of a specified time period for predetermined cell locations. Detect when a subscriber connects and retrieve operatorprovided location ID. If MNO provides network policies, decide on a level of traffic management based on policy. or If MNO monitors UXI: If no previous UXI exists, track UXI and decide when to apply traffic management Serve videos with increased traffic management for all new subscribers of the cell at the specified time and for new video sessions by current subscribers If existing UXI is unsatisfactory, decide traffic management based on policy If MNO tracks UXI and RAN input, apply video optimization for subscribers with unsatisfactory UXI Dynamic Congestion Scenario At the cell level, network traffic changes by the millisecond based on instantaneous loads from many different subscribers. To respond to this dynamic network traffic, Adaptive Traffic Management assesses real-time network conditions, monitoring inputs from network probes or other RAN elements from different RANs that may indicate congestion. In addition, Adaptive Traffic Management can identify the condition of a cell by detecting Dynamic Bandwidth Shaping (DBS) across subscriber sessions within the cell.
11 11 If dynamic RAN inputs or application of DBS indicate congestion, Adaptive Traffic Management applies traffic management to all new subscribers in the problem cell. Consider a scenario in which a subscriber is watching a video and many other subscribers in that same cell begin watching videos or begin using their smartphone applications for social networking or web browsing. The additional network traffic generated by the other subscribers negatively impacts the first subscriber s video connection and stalls his video. In this scenario, if the subscriber requests another video within a specified period of time, Adaptive Traffic Management serves a video with increased video compression, reducing the load on the cell. Table 3. Adaptive Traffic Management Based on Dynamic RAN Input Scenario Condition Detect Decide React Subscriber is watching a video. Other subscribers in the same network location (cell) begin watching videos. Detect network probe or other RAN element indicating congestion or Detect prevalence of DBS application to videos for subscribers within a cell. If the MNO provides network policies, decide on the right level of video optimization based on policy. or If the MNO uses UXI, and no previous UXI exists, track UXI If existing UXI is unsatisfactory, decide on the right level of traffic management based on policy Serve videos with increased content optimization for all new subscribers of the cell and for video sessions by current subscribers. If the MNO tracks UXI and RAN input or the prevalence of DBS, apply content optimization for subscribers with either unsatisfactory UXI and RAN inputs or DBS tracking that indicates congestion. Adaptive Traffic Management Based on a Specific Application Video consumes a major percentage of mobile bandwidth, but other applications like peer-to-peer file sharing, large s, or auto-update downloads can also consume significant bandwidth. Adaptive Traffic Management works for these applications as well. Imagine a scenario in which a subscriber is watching a video and then initiates a P2P session and downloads a large file. The additional network traffic generated by the subscriber s action negatively impacts the subscriber s video connection and risks stalling the video. In this scenario, rather than apply traffic management to the video stream, Adaptive Traffic Management may apply bandwidth shaping to the P2P session and file download, reducing the load on the cell and allowing the video to be viewed without having to resort to content-optimization techniques.
12 12 Adaptive Traffic Management based on a Specific Application Table 4. Adaptive Traffic Management Based on a Specific Application Scenario Condition Detect Decide React Subscriber is watching a video. Subscriber initiates a P2P file sharing session and downloads a large file. Detect changes in UXI across all subscribers. If UXI is not satisfactory or subscriber is about to experience a video stall, decide to apply P2P bandwidth shaping and/or file download throttling based on rate plan or network policy. Apply bandwidth shaping to P2P session and file download sessions. Adaptive Traffic Management Based on a Global View of Subscribers To address the growing challenge of OTT streaming and progressive download video and to counteract the impact of congestion, MNOs can deploy Adaptive Traffic Management with a global view, applying lossless optimization and policy control at the content and application layers. This allows MNOs to: Control encrypted and non-encrypted streaming video through Streaming Policy Control (SPC) Minimize waste of progressive download videos through JIT video delivery Shift automated software downloads away from peak hours through software download policy control Apply bandwidth shaping to P2P traffic and file downloads to limit their impact during times of network congestion
13 13 Adaptive Traffic Management dynamically measures the UXI for all subscribers in real time. When the UXI begins to deteriorate for a specific subscriber, Adaptive Traffic Management invokes the appropriate traffic management technique for other subscribers for example, SPC for streaming video or bandwidth shaping for P2P traffic while applying software download policy control for all subscribers. Envision a scenario in which a high-value subscriber is watching a video and the traffic on the subscriber s cell increases slightly, perhaps because other subscribers begin to watch streaming videos or have begun to engage in P2P file transfers. The additional network traffic generated by other subscribers impacts the first high-value subscriber. In this scenario, Adaptive Traffic Management applies SPC to streaming video as needed and throttles P2P traffic to all subscribers to maintain QoE. Adaptive Traffic Management based on a Global View Table 5. Adaptive Traffic Management Based on a Global View Scenario Condition Detect Decide React High-value subscriber is watching an HTTP progressive download video from a subscription service. Traffic on high-value subscriber s cell increases when other subscribers initiate P2P sessions or watch streaming videos. Detect changes in UXI across all subscribers. If UXI is not satisfactory for high-value subscriber based on policy, rate plan or subscriber class: Decide how aggressively to apply P2P bandwidth shaping and SPC for streaming traffic. Decide which subscribers qualify for improvement Apply SPC to specific subscribers to control their video resolution. Apply bandwidth shaping to P2P sessions for specific subscribers. Detect changes to UXI for all subscribers, especially for highvalue subscriber
14 14 Deploying Adaptive Traffic Management To maintain the highest possible QoE for all subscribers, MNOs can easily deploy Adaptive Traffic Management in both existing 3G networks and new 4G networks. 3G Networks Although 3G networks were a welcome improvement when first introduced, better technologies 3GPP LTE and true 4G networks have since been developed. In light of this, MNOs may not want to invest in additional 3G capacity, preferring that new capital expenditures go to more modern 4G networks. However, with the rapid uptake of smartphones and the explosion of high-bandwidth media music and video 3G networks can reach capacity during peak hours. Applying Adaptive Traffic Management can help increase the effective capacity of existing 3G networks without requiring significant additional investment in 3G network equipment. By applying the right traffic management techniques for the subscribers who need it, at the right time, Adaptive Traffic Management ensures that 3G subscribers continue to receive high QoE. LTE/4G Networks With new LTE and LTE-Advanced networks, MNOs have an opportunity to segment their subscriber base, leveraging subscriber intelligence to create new plans that can deliver deterministic, high QoE. Operators can add new models based on subscriber application and content interests and QoE to their existing quota-based plans. Although these networks seem to have sufficient capacity today, with the exponential growth of high-bandwidth traffic they will become capacity constrained. To forestall this inevitability, forward-thinking MNOs will start adding controls to LTE and 4G traffic by applying Adaptive Traffic Management as soon as possible to help manage traffic growth. In addition, certain applications and OTT video are designed to be optimized to maintain high QoE for a single subscriber: video servers transmit data as fast as the network permits, consuming all available bandwidth for the duration of the video delivery at the expense of other subscribers and applications. By applying Adaptive Traffic Management, MNOs can start controlling the bandwidth consumed by OTT video traffic before there is a problem and before other subscribers QoE suffers. Conclusion By applying the right traffic management techniques to the right subscribers at the right time, based on policy inputs, subscriber QoE and RAN conditions, Adaptive Traffic Management can help MNOs deliver the highest possible QoE to all subscribers, extend the capacity of existing 3G networks without additional capital expenditures, control OTT traffic in new LTE or 4G networks, and increase revenue.
15 15 Abbreviations Abbreviation Codec DBS DPI HTTP JIT LTE MNO MOS MPEG OTT P2P QAT QoE QoS RAN RNC RTMP SLA SPC TCP UXI Meaning Encoder/decoder Dynamic Bandwidth Shaping Deep packet inspection Hypertext transport protocol Just-in-time Long-term evolution Mobile network operator Mean opinion score Moving Picture Experts Group Over-the-top Peer-to-peer Quality Aware Transcoding Quality of Experience Quality of Service Radio access network Radio Network Controller Real-time messaging protocol Service level agreement Streaming policy control Transmission control protocol User experience index
16 16 Appendix: Underlying Techniques of Adaptive Traffic Management Application Detection and Control Application detection and control examines the application payload of a packet or traffic stream and makes traffic management decisions for that data based on its content. Application detection classifies traffic into categories: video streaming, web traffic, video chat, P2P and file downloads, social network, and the like. It can classify traffic at the subscriber level and at a global level. Control can be applied on a global or per-subscriber basis. Content-Layer Control Unlike typical deep packet inspection (DPI) tools, when making traffic management decisions Adaptive Traffic Management uses information not only from the protocol layers but also from the content layers. For content-layer traffic, it uses Citrix ByteMobile lossless optimization capabilities, whereby web and video content are subjected to standards-based techniques that reduce the amount of data across the network while the original content is generated or recreated at the subscriber s smartphone, tablet or laptop. These lossless techniques do not impact the original video or web object (e.g., image). Content-layer control, which works primarily at the content layer but can also operate at the transport layer, consists of the following: Policy control, to address streaming video, automated software updates and application throttling Lossless video optimization, consisting of video pacing using JIT video delivery Lossless web optimization, such as lossless image compression and page compression with protocol and transport optimization Citrix ByteMobile Content-Layer Optimization Content-layer optimization achieves a balance between compression and quality of subscriber experience by performing Quality Aware Transcoding (QAT), detecting the quality of the web object or video, and applying lossy compression at the necessary level: For video, QAT may remove any inefficiency in the original encoding of the video by replacing inefficient codecs, tuning the encoding profile for higher efficiency, removing video frames that cannot be seen, or eliminating redundant frame data. QAT can be configured to deliver a range of video compression, from visually lossless to slightly degraded. For images, QAT uses informed transcoding techniques (for example, JPEG compression with varying quality factors) to balance the quality of the image while reducing the amount of network bandwidth consumed in transmitting it.
17 17 Adaptive Traffic Management can apply QAT dynamically, based on the changing speed of the subscriber s connection. For example, if a subscriber s connection speed slows, the system automatically increases the compression level of the video to reduce required bandwidth and prevent stalling. This type of dynamic QAT is referred to as Dynamic Bandwidth Shaping (DBS). In all cases, Adaptive Traffic Management delivers videos and web objects at the appropriate levels of compression based on policy inputs, subscriber QoE and the condition of the RAN. In this way, Adaptive Traffic Management helps maintain the highest possible QoE for all subscribers and expands network capacity by up to 50%. Content Caching Content caching involves storing streaming and progressive download video and web content relatively close to the subscriber, helping MNOs eliminate the impact of content server delays, decrease the volume of internet traffic and reduce the required hardware footprint. Although caching is simple in principle, to be effective it must be: Transparent: Content server changes must not impair caching, and caching must not impair content server functions, such as traffic measurement or advertising Effective: Popular videos should be cached using algorithms that maximize the cache hit rate while balancing storage requirements Efficient: Videos should be saved to the cache only during subscriber downloads, avoiding the creation of additional traffic on the internet backhaul for a video that few subscribers view in its entirety In this way, caching accelerates video content delivery to ensure smooth viewing without interruptions or buffering delays. Serving videos from the cache eliminates the several seconds of delay introduced by the content server and internet peering links. It also reduces bandwidth use and costs of peering and transit links, as video downloads are served more frequently by the local cache rather than by out-ofnetwork servers.