White Paper Why Is DPI Essential for Wireless Networks? By: Manish Singh, CTO Overview The biggest challenge network operators continue to face is how to keep up with seemingly insatiable demand for mobile broadband. The basic paradigm in mobile wireless networks has been to have a shared data channel a big fat bit pipe in every cell site whereby all the data users in that cell site share the channel capacity. Unlike voice, for which dedicated channels are allocated, for the most part data channels (both uplink and downlink) are shared. Theoretically the paradigm works well for data networks because data is considered to be bursty in nature, meaning the shared resource is typically available because each user s relative consumption of it is generally quite small and fast. CONTENTS Business Case for Mobile Broadband pg. 2 Deep Packet Inspection (DPI) pg. 2 Deploying DPI pg. 4 Adaptive Traffic Shaping Tying RAN Loading to Traffic Shaping pg. 5 Conclusion pg. 5 References pg. 6
2 However, the nature of data traffic itself is changing. For instance, a simple Internet video download from YouTube sucks bandwidth out the network for sustained durations. Peer-to-peer (P2P) traffic is the root cause of congestion in Internet Service Providers (ISPs ) wireline networks, and mobile networks are just not ready to handle the traffic volumes generated by P2P. As such, strong uptake rates of wireless data cards, USB dongles and emerging netbooks all present a real threat: P2P traffic migrating from wireline networks to mobile broadband networks. Streaming traffic, like Internet radio, is another example of data traffic that breaks the shared channel paradigm. And last but not the least, Web site content is itself becoming heavier, for the average Web page size has tripled since 2003 as both the number of contained objects and the size of those objects continue to increase. 2 Because of these facts, data can no longer be considered only bursty in nature. Figure 1. Channel Structures in a 3G HSPA Cell To meet the growing mobile broadband data demand, two main techniques are available to wireless network operators for adding capacity: Increasing Spectral Efficiency: Measured in bits/sec/ Hz. Long Term Evolution (LTE) promises four times improvement over its predecessor, WCDMA. Splitting Cells: Adding new, smaller cell sites thereby reducing subscriber density in a given cell site. Growth in data rates far outstrips the improvements in spectral efficiency, however. For example, LTE, which improves spectral efficiency by four times, will take anywhere from five to ten years for complete network rollouts and broad consumer adoption. In the meantime, mobile broadband operators are seeing traffic growth ranging from 400 percent to 700 percent annual and so spectral efficiency improvements are just not enough. Similarly, adding new cell sites is a costly business, and operators need to find new ways to better monetize their current network infrastructure. All of these factors help underscore the business case for mobile broadband. Figure 2. Content is Becoming Heavier
3 Business Case for Mobile Broadband Unlike for voice, mobile broadband s business models are complex. Continuous innovation on the Internet is driving new applications and content that captures consumers interests. The reality, however, is that mobile operators are not the best innovators for new data applications and content and there is no better example to prove this claim than the fact that not a single mobile operator Web site shows up in the top 100 visited Web sites. From Yahoo to Flickr to YouTube, data traffic is dominated by Internet application and content providers. The challenge for mobile operators, then, is whether they should be relegated to being mere bit pipe providers or whether there are monetization opportunities beyond just providing seamless connectivity. One key constraint on wireless networks is that they are inherently spectrum limited; throwing more spectrum at the problem is simply not an option for wireless networks (unlike the ability of wireline network operators to throw more fiber at bandwidth problems). At times, Radio Access Networks (RAN) are also capacity-limited by backhaul bottlenecks between NodeBs and RNCs. Thus operators need to find ways to effectively manage growing data traffic in their networks, and this must be done in ways where network monetization is maximized. Deep Packet Inspection (DPI) DPI is an emerging technology area which is enabling new innovative applications for data networks. Simply put, DPI technology is able to open up every IP packet and inspect the contents of the packet all the way up to Layer 7 of the OSI stack, all in real-time. In today s environment, traffic management schemes using IP port numbers are just not enough because application developers embed all kinds of content deeper inside the packets. Fortunately, DPI enables mobile operators to develop a thorough understanding of the types of Figure 3. Voice vs. Data: Old vs. New & Complex Business Models Figure 4. Throwing More Spectrum to Meet Growing Data Traffic Demand Is Not an Option traffic flowing in their networks and, based on that information, they can then prioritize, rate limit, and in extreme scenarios, block different types of traffic. Operators can implement effective policies and pricing plans for data traffic and enforce them using DPI for data traffic management. As examples, using DPI, an operator can enforce a policy of rate-limiting P2P traffic, or prioritize enterprise application traffic over Internet video-streaming traffic. The possibilities are boundless.
4 DPI also enables network operators to implement intelligent network traffic offloading. One of the growing interest areas for mobile operators is to find ways to offload Internet traffic directly to the Internet and not carry that traffic over their core networks. However, the devil lies in the details when attempting to do effective traffic offloading, for two RTP streams might look very similar and if the operator is offering VoIP services, simply offloading all RTP traffic is not good enough. With DPI, operators can intelligently analyze the data streams in real-time and apply sophisticated traffic offloading schemes. Most importantly, DPI creates new monetization opportunities for mobile network operators. With DPI, operators can monetize both sides of content delivery (a) with consumers, by offering tiered SLAs and prioritized traffic plans and (b) with content providers, by offering tiered Internet-like services and creating fast paths for them. However, operators must take these steps in a very careful manner, as these changes often strike right in the heart of net neutrality. Many believe that the concept of net neutrality for wireless networks is a fallacy, as throwing more fiber at the problem is simply not enough because these networks are inherently spectrum limited. DPI enables network operators to convert bit pipes to smart pipes, which is ultimately a win-win-win scenario for subscribers, content providers and operators alike. IP networks are inherently prone to security challenges. From Distributed Denial of Service (DDoS) attacks to intrusion to viruses, mobile broadband networks like wireline networks need to be secured from the latest emerging security threats. Fortunately, DPI enables wireless network operators to deploy comprehensive security solutions in the network, protecting them from zero-day attacks to existing well-known viruses. As with wireline networks, lapses in security in mobile broadband networks are just not an option and DPI holds the key for creating clean pipes. Figure 5. Smart Pipes = Bit Pipes + DPI Figure 6. Clean Pipes = Bit Pipes + DPI Deploying DPI DPI deployments first started in wireline networks as ISPs started implementing solutions for traffic management and security. These deployments were mostly bump-in-the-wire whereby a DPI box was put inline in the data traffic path and the box would inspect traffic and thus enable traffic management and security applications. While effective, the bumpin-the-wire deployment scenario requires additional boxes in networks which inherently increase complexity and associated OpEx and CapEx.
5 Wireless network operators who have already deployed 3G HSPA networks have no choice but to adopt the bump-in-the-wire deployment schemes. However, greenfield HSPA deployments can have DPI capabilities embedded inside wireless gateways SGSN, GGSN, Femto Gateway, etc., and DPI-enabled gateways allow network operators to reduce their network complexity while achieving the desired functionality. What s more, DPI is essential for LTE, or 4G wireless networks. LTE will be the first true All-IP network with a converged Evolved Packet Core (EPC). A converged core delivers significant benefits to a mobile operator because its core network is simplified, which substantially reduces OpEx and CapEx. On the other hand, an All-IP core brings with it the challenge of traffic management, Quality of Service (QoS) and security. In the future, as operators carry both voice and data traffic over their converged EPC, DPI will be essential for prioritizing real-time voice traffic over data and other less time-sensitive traffic. For LTE networks, DPI functions must be embedded in the Serving Gateway for effective traffic management, QoS and security. Adaptive Traffic Shaping Tying RAN Loading to Traffic Shaping Wireless networks provide ubiquitous mobility and coverage by deploying multiple cell sites over a given geographic area. However, not all cell sites are loaded equally, for such loading depends on location, time of day, number of users, etc. Furthermore, cell loading changes over time, as an airport cell site, for instance, might be heavily congested during business days but might have full capacity available throughout the night. Traffic shaping in wireline networks is predominantly done by classifying traffic and then applying policies. In the case of wireless networks, tying cell site loading to traffic shaping provides additional compelling advantages to network operators. Extending the example above, an operator could prioritize enterprise traffic over Internet video or P2P traffic at the airport cell site when it is heavily loaded; conversely the operator might allow all traffic in the evening hours when plenty of spare capacity is available at the same Figure 7. DPI Deployment Options cell site. Such a traffic management scheme enables operators to offer tiered SLAs to their subscriber base, thereby increasing satisfaction and creating opportunities for additional revenue streams. Radisys patented Adaptive Traffic Shaping architecture solution ties real-time cell loading to DPI-based traffic management. The solution monitors cell loading in realtime and uses this information to dynamically shape the traffic using DPI, thereby enabling network operators to maximize the monetization of their network assets. Conclusion Mobile broadband adoption is today s growth engine for wireless network operators. However, mobile networks are inherently capacity-limited because of finite wireless spectrum. While technology innovation is continuously improving spectral efficiency, such capacity gains are tiny in comparison to the current and projected massive growth in data traffic. Effective traffic management in mobile broadband networks is essential, and DPI holds the key. Adaptive traffic shaping takes DPI-based traffic management solutions to the next level by tying real-time cell loading conditions to the traffic shaper. DPI enables network operators to explore new business models by converting ordinary bit pipes into smart pipes and creating new revenue streams such as custom service plans and targeted ad insertion.
6 Likewise, securing mobile broadband networks is essential and DPI also enables operators to deliver clean pipes to their subscribers. Poor QoS and lapses in security are imminent threats to mobile broadband adoption, particularly in an All-IP environment like LTE. If network operators are to capitalize on the tremendous growth opportunity of mobile broadband, as well as meet lawful intercept requirements required by national regulations, they now know exactly where solutions exist for their problems: DPI. References 1 CommsDay International Edition, Jan 27, 2009. 2 Domenech 2007, Gomez 2008 Domenech, J., Pont, A., Sahuquillo, J., and J. Gil, A user-focused evaluation of web prefetching algorithms, Computer Communications 30, no. 10 (2007): 2213-2224. In 1995 there were 2.3 average objects per page and 25.7 in 2003 (average of two traces). Flinn, D., and B. Betcher of Gomez.com, Re: latest top 1000 website data? e-mail to author, Jan. 8, 2008. As of January 2008, the average top 1000 home pages was 312K in total file size, referencing 49.92 total objects. http://www.websiteoptimization.com/speed/tweak/ average-web-page/ Corporate Headquarters 5435 NE Dawson Creek Drive Hillsboro, OR 97124 USA 503-615-1100 Fax 503-615-1121 Toll-Free: 800-950-0044 www.radisys.com info@radisys.com 2011 Radisys Corporation. Radisys, Trillium, Continuous Computing and Convedia are registered trademarks of Radisys Corporation. *All other trademarks are the properties of their respective owners. September 2011