際際滷

際際滷Share a Scribd company logo

Latency Considerations in LTE: Implications to Security Gateway

T
Terry Young

This white paper discusses latency considerations for LTE and LTE-Advanced networks. Latency requirements are becoming more stringent over time. LTE-A targets latency of 10ms or less, with 1ms or less required for the X2 interface to support new optimization techniques. Higher latency can negatively impact user experience through slower page loads and reduced throughput. It can also result in lost revenue for online businesses. Any network element must minimize its contribution to overall latency in order to meet budgets. Low-latency solutions like the Stoke Security eXchange are important for meeting stringent LTE-A requirements.

1 of 11
Downloaded 21 times
WHITE PAPER Latency Considerations in LTE Implications to Security Gateway September 2014
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 2 Contents Executive Summary ......................................................................................... 3 Latency - LTE's New Performance Metric ....................................................... 4 Calculating Total Latency............................................................................................................ 4 Stringent New Requirements with LTE-A ........................................................................... 5 Microsecond Performance for X2 ............................................................................................ 6 X2 Delay and User Throughput ............................................................................................ 7 Latency Impacts the User Experience ................................................................................... 8 Lower Latency Improves Page Load Times .................................................................... 8 Estimated Sales Impact ........................................................................................................... 9 M2M and On-Line Gaming ..................................................................................................... 9 Implications for Security Gateway .................................................................................. 10 Conclusions .................................................................................................................................... 11 Stoke速 Security eXchange ............................................................................ 11
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 3 Executive Summary Network delay or latency is critical in todays mobile broadband where both Internet-based businesses and user expect network response will be close to instantaneous. Excess latency can have a profound effect on user experiencefrom excess delay during a simple phone conversation, reducing throughput at edge of cell coverage areas by reducing effectiveness of RAN optimization techniques, to slow-loading webpages and delays with streaming video. Response delays negatively impact revenue. In financial institutions, low latency networks have become a competitive advantage where even a few extra microseconds, can enable trades to execute ahead of the competition. The direct correlation between delay and revenue in the web browsing experience is well documented. Amazon famously claimed that every 100 millisecond reduction in delay led to a one percent increase in sales. Google also stated that for every half second delay, it saw a 20 percent reduction in traffic. For LTE network operators, control of latency is growing in importance as both an operational and business issue. Low latency is not only critical to maintaining the quality user experience (and therefore, the operator competitive advantage) of growing social, M2M, and real-time services, but latency reduction is fundamental to meeting the capacity expectations of LTE-A, where latency budgets will be cut in half and X2 will need to perform at microsecond speed. Total network latency is the sum of delay from all the network components, including air interface, the processing, switching, and queuing of all network elements (core and RAN) along the path, and the propagation delay in the links. With ever tightening latency expectations, the relative contribution of any individual network element, such as a security gateway, must be minimized. For example, when latency budgets were targeting 150ms, a network node providing packet processing at 250亮s was only adding 0.17% to the budget. However, in LTE-A, with latency targets slashed to 10ms, that same network node will consume almost 15x more of the budget. More important, when placed on the S1 with a target of only 1ms, 250 亮s is 25% of the entire S1 latency allocation, and endangers meeting the microsecond latency needed at the X2. Clearly, operators need to apply stringent latency requirements for all network nodes, when designing LTE and LTE-A networks. STOKE速 solutions are purpose built from the ground up for today's mobile broadband environment to solve critical, performance impacting problems for mobile network operators. Stoke Security eXchange is a carrier grade, field proven solution ideal for these requirements since it introduces <30 microseconds of latency, supports creation of multiple VLANs and line rate/10Gbps throughput for small packet sizes of 96 bytes.
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 4 Latency - LTE's New Performance Metric Latency and throughput are the essential factors in network performance and collectively they define the speed of a network. Whereas throughput is the quantity of data that can pass from source to destination in a specific time, round trip time (RTT) latency is the time it takes for a single data transaction to occur, meaning the time it takes for the packet of data to travel to and from the destination, back to the source. Latency is often considered to be even more important than speed in determining quality user experience: Network latency, perhaps more so than average downlink speeds alone, can affect users' experience, especially for real-time services like video calling, VoIP and even gaming applications. Source: Light Reading1 As consumer and businesses increasingly merge user experience across multiple devices and networks, mobile networks must also be engineered to minimize delay. Calculating Total Latency The total latency budget is measured in milliseconds (ms), but is comprised of individual network elements and interfaces, which individually may only add microseconds (亮s), but must be added together to calculate total end-to-end latency. Latency must be carefully managed and measured. Latency includes delay from propagation, buffering and queuing, transmission, and signal processing that is introduced at every link and network element through which a packet travels. A primary design objective for any individual network element (such as a security gateway) should be to minimize its latency contribution in order to stay within the overall latency budget prescribed for the link. As mobile technology has evolved, the latency targets have become increasingly stringent. The figure following shows the progression of roundtrip times from GSM/Edge to LTE, as defined by 3GPP. 1 Light Reading, LTE, A latent problem
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 5 Figure 1. Evolution of Mobile Network Latency Targets2 As shown in Figure 1, for LTE networks, the 3GPP target latency budget for the user plane latency (radio and core) is about 20 ms, excluding the backhaul transport network. The backhaul network (RAN edge- Core edge, including security gateway) would add an additional 10 ms.3 Stringent New Requirements with LTE-A In order to support the predicted growth, the 3GPP has developed LTE-Advanced Release 10, a major enhancement of the LTE standard deployed in many mobile networks today. The new technology targets peak data rates up to 1 Gbps and introduces new RAN concepts with the ultimate goal of designing a system that is drastically enhanced in both cell capacity and coverage. In LTE Advanced, latency targets are reduced substantially to 10 ms, allocated as shown following. Figure 2. Target Latency Budget for LTE-A4 2 Nokia Siemens Networks, LTE-capable transport: A quality user experience demands an end-to-end approach 3 Qualcomm, Latency in HSPA Data Networks, July 2013.
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 6 Microsecond Performance for X2 To maintain performance requirements in LTE-A, the latency on the X2 links are especially critical: In order to meet stringent latency requirements of less than 1 millisecond, the physical and logical path of the X2 interface needs to be as short as possible. 5 In LTE-Advanced, new RAN optimization techniques introduced in LTE-A impose critical performance demands on the X2 interface, requiring very short latencies of <1 millisecond across the backhaul network. This is further described: Inter-cell interference coordination (ICIC) and coordinated multi-point (CoMP) transmission are two techniques defined in LTE-Advanced specifications that target a better user experience at the cell edge. ICIC is limiting cross-talk by coordinating spectrum allocation across multiple cells. CoMP allows multiple base stations to simultaneously serve a user device and increase the receive power level and, therefore, capacity. Both synchronization techniques are implemented over the X2 interface and require very short latencies of <1 millisecond across the backhaul network to achieve real-time coordination between base stations. In addition to providing low-latency connectivity, base station clocks need to be in phase to enable proper operation of ICIC and CoMP. This leads to the requirement for highly accurate phase or time-of-day synchronization. Most 3GPP base station clocks are currently synchronized on frequency only, since accurate phase synchronization was not a requirement up to now. The new LTE-Advanced functions, however, require base stations to be in phase with an accuracy of 500 nanoseconds to efficiently operate ICIC and CoMP. This is nearly impossible to achieve without active time distribution over the X2 interface. Mobile operators will also introduce LTE TDD radio interfaces operating in unpaired spectrum. Many operators already have acquired unpaired spectrum, since TDD provides more flexible scaling of the up- and down-link capacity and has additional benefits to the overall architecture of the radio access network. Like ICIC and CoMP, LTE TDD also requires phase alignment of neighboring base stations over the X2 interface in addition to the traditional frequency synchronization used in mobile networks today. 4 LOLA, Presentation of WP2 Scenarios and Target System Architectures 5 RCR Wireless Readers Forum, New Backhaul Challenges are emerging with LTE Advanced, July 15, 2013.
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 7 These techniques mandate tighter coordination between base stations and places specific performance requirements for the X2 interface with respect to capacity, latency and synchronization. Supporting these requirements is not trivial since most mobile backhaul architecture on which X2 interface is transported follows a hub-and-spoke design where traffic distribution and re-direction is performed at the security gateway in the distribution layer of the backhaul network. X2 Delay and User Throughput Latency on the X2 interface reduces the benefit of co-ordination schemes. In the figure following, X2 latency as low as 5ms reduced cell edge and median user throughputs by over 20% in this example of a Joint Transmission CoMP scheme.6 This loss of throughput impairs the user experience and reduces network efficiency. Figure 3. Impact of X2 delay on user throughput with CoMP scheme 7 Simpler schemes may not be impacted as much, but this shows that the sophisticated CoMP schemes require very low X2 latencies to achieve their full potential and delivering the anticipated enhancements in speed, coverage and overall quality of experience to the end subscriber. The table below uses the 5ms and 1ms figures above and calculates the throughput loss for 200 亮s. X2 Delay 5 ms 1 ms 200 亮s Throughput Loss (20%) (5%) (1.0%) Figure 4. Loss of Spectral Efficiency from X2 Delay. 6 Qualcomm, Backhaul Requirements for Centralized and Distributed Cooperation Techniques, 8 July 2010. 7 Centralized Scheduling for Joint-Transmission Coordinated Multi-Point in LTE-Advanced, S. Brueck, L. Zhao, J. Giese, M. A. Awais, proc. ITG/IEE Workshop on Smart Antennas, Bremen (WSA'10), Germany, February 2010
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 8 Latency Impacts the User Experience High latency causes noticeable delays in, for example, downloading Web pages or when using latency sensitive applications such as interactive games, VoLTE, M2M, and streaming media.8 Lower Latency Improves Page Load Times Analysis by Google shows a relationship between page load time and round trip latency. Their analysis shows reductions of every 20 milliseconds in round trip latency reduces page load time by 7-15%.9 The figures following compare the impact on page load time from two different performance characteristics - bandwidth improvements and round-trip-time (RTT). There are diminishing returns as the bandwidth gets higher, but not for improvements in round trip time (latency). Google concluded that, unlike improvements to bandwidth, reducing the round trip time always helps the overall page load time. Figure 5. Comparison of impact on page load times RTT and Bandwidth The linear relationship shown in the charts provided by Google suggest that microsecond latency improvements would also improve the page load times and therefor have some impact on the user experience. Both Amazon and Google have confirmed that in e-commerce applications, milliseconds of higher page load times, which is not consciously perceivable to the user, can still have an impact on user behavior and experience and can directly impact revenue. 8 Nokia Siemens, Latency The impact of latency on application performance 9 Google Blog, More Bandwidth Doesnt Matter (much ) , April 2010
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 9 Any increase in the time it takes Google to return a search result causes the number of search queries to fall. Even very small differences in results speed impact query volume.10 Experiments at Amazon have revealed similar results: every 100 ms increase in load time of Amazon.com decreased sales by one percent.11 Estimated Sales Impact12 Lower latency, even in microseconds, improves page load times, which in turn can positively impact revenue for mobile operators and enterprise customers. To illustrate the potential impact on sales from less efficient network nodes, the table below shows the potential impact of additional 200 亮s latency (page load times) on an on-line enterprise with $14.8B in annual sales, using the 100 ms/1% sales decrease provided by Amazon. As shown in Figure 6, the estimated impact on sales for each additional 200 亮s of page load delay could be as high as $300k annually. Example - Sales Lost Per 100 ms Per ms 200 亮s 2007 Annual Sales : $14,800 M Negative Impact on Sales (1.0%) (.01%) (0.002%) Sales Lost ($148 M) ($1.48 M) ($0.3 M) Figure 6. Estimated sales lost for 200 microseconds delay. M2M and On-Line Gaming Two emerging, massive applications also require low latency: High Performance On-line Gaming M2M, Sensory applications Ericsson expects that in 2020 there will be 50 billion devices connected and available to be used in various existing and new applications. The figures below compare the latency requirements of new and existing applications and provide latency targets of a few representative applications. 10 CNET News.com/ZDNET.com: Google卒s Marissa Mayer: Speed wins by Dan Farber, November 9, 2006. 11 Nokia Siemens Networks: Latency: The impact of latency on application performance, 2009. 12 In order to estimate the impact of an additional 200亮 of latency, a linear relationship with the 100ms-level data available is assumed.
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 10 Figure 7. New High Growth Applications require Low Latency13 M2M and Gaming Application Latency Target On-Line Gaming 80 ms (U-Plane, 1 way) Gaming Sports Events 25 ms (U-Plane) Sensor-Based Alarms 2-12 ms (1 way, U+C plane) Figure 8. Example latency targets of M2M and Gaming applications.14 Implications for Security Gateway The necessity to meet the <1 millisecond budget of the X2 interface in turn places strict requirement to minimize IPsec and packet routing latency in the security gateway, which is the key nodal element responsible for security and routing of control and data plane traffic in X2 interface. The entire S1 interface budget is only 1 ms (1,000 亮s) each way). Therefore, the example node referenced earlier that contributed 250 亮s of latency, when placed on the S1 would be consuming 25% of the entire backhaul latency budget. Each additional 200 亮s delay from a security gateway would cause a loss of 1% throughput. In contrast, the STOKE速 Security eXchange is engineered to contribute only 30 亮s or 3% of the 1 ms latency budget. 13 Eurocom, Achieving LOw-Latency in Wireless Communications 14 Eurocom: Presentation of WP2 Scenarios and Target System Architectures
Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 11 Conclusions Network latency, even more than download speeds, directly impacts the user experience and bottom line revenue for on-line businesses. In high frequency financial market trading, microsecond improvements are considered a competitive advantage. As low latency applications grow in importance and e-commerce increasing moves to mobile platforms, operators will need to carefully manage and measure latency budgets to maintain their own competitive advantage. The latency contribution of all individual network elements, including the security gateway, must be carefully calculated. In LTE-A and especially for the X2 interface where the latency targets are drastically reduced, an additional 200 亮s delay is a significant difference. Stoke速 Security eXchange Operators are challenged to integrate networks and technologies smoothly, sustain a quality user experience with high network performance, and still keep service delivery costs low. Stoke solutions are purpose built from the ground up for today's mobile broadband environment to solve critical, performance impacting problems for mobile network operators. Stoke innovative design and patent pending technologies enable cost effective, concurrent operation of critical functions while maintaining line-rate, high performance throughput. STOKE Security eXchange is a carrier grade, field proven solution ideal for these requirements since it introduces < 30 microseconds of latency, supports creation of multiple VLANs and line rate/10Gbps throughput for small packet sizes of 96 bytes.

More Related Content

Latency Considerations in LTE: Implications to Security Gateway

  • 1. WHITE PAPER Latency Considerations in LTE Implications to Security Gateway September 2014
  • 2. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 2 Contents Executive Summary ......................................................................................... 3 Latency - LTE's New Performance Metric ....................................................... 4 Calculating Total Latency............................................................................................................ 4 Stringent New Requirements with LTE-A ........................................................................... 5 Microsecond Performance for X2 ............................................................................................ 6 X2 Delay and User Throughput ............................................................................................ 7 Latency Impacts the User Experience ................................................................................... 8 Lower Latency Improves Page Load Times .................................................................... 8 Estimated Sales Impact ........................................................................................................... 9 M2M and On-Line Gaming ..................................................................................................... 9 Implications for Security Gateway .................................................................................. 10 Conclusions .................................................................................................................................... 11 Stoke速 Security eXchange ............................................................................ 11
  • 3. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 3 Executive Summary Network delay or latency is critical in todays mobile broadband where both Internet-based businesses and user expect network response will be close to instantaneous. Excess latency can have a profound effect on user experiencefrom excess delay during a simple phone conversation, reducing throughput at edge of cell coverage areas by reducing effectiveness of RAN optimization techniques, to slow-loading webpages and delays with streaming video. Response delays negatively impact revenue. In financial institutions, low latency networks have become a competitive advantage where even a few extra microseconds, can enable trades to execute ahead of the competition. The direct correlation between delay and revenue in the web browsing experience is well documented. Amazon famously claimed that every 100 millisecond reduction in delay led to a one percent increase in sales. Google also stated that for every half second delay, it saw a 20 percent reduction in traffic. For LTE network operators, control of latency is growing in importance as both an operational and business issue. Low latency is not only critical to maintaining the quality user experience (and therefore, the operator competitive advantage) of growing social, M2M, and real-time services, but latency reduction is fundamental to meeting the capacity expectations of LTE-A, where latency budgets will be cut in half and X2 will need to perform at microsecond speed. Total network latency is the sum of delay from all the network components, including air interface, the processing, switching, and queuing of all network elements (core and RAN) along the path, and the propagation delay in the links. With ever tightening latency expectations, the relative contribution of any individual network element, such as a security gateway, must be minimized. For example, when latency budgets were targeting 150ms, a network node providing packet processing at 250亮s was only adding 0.17% to the budget. However, in LTE-A, with latency targets slashed to 10ms, that same network node will consume almost 15x more of the budget. More important, when placed on the S1 with a target of only 1ms, 250 亮s is 25% of the entire S1 latency allocation, and endangers meeting the microsecond latency needed at the X2. Clearly, operators need to apply stringent latency requirements for all network nodes, when designing LTE and LTE-A networks. STOKE速 solutions are purpose built from the ground up for today's mobile broadband environment to solve critical, performance impacting problems for mobile network operators. Stoke Security eXchange is a carrier grade, field proven solution ideal for these requirements since it introduces <30 microseconds of latency, supports creation of multiple VLANs and line rate/10Gbps throughput for small packet sizes of 96 bytes.
  • 4. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 4 Latency - LTE's New Performance Metric Latency and throughput are the essential factors in network performance and collectively they define the speed of a network. Whereas throughput is the quantity of data that can pass from source to destination in a specific time, round trip time (RTT) latency is the time it takes for a single data transaction to occur, meaning the time it takes for the packet of data to travel to and from the destination, back to the source. Latency is often considered to be even more important than speed in determining quality user experience: Network latency, perhaps more so than average downlink speeds alone, can affect users' experience, especially for real-time services like video calling, VoIP and even gaming applications. Source: Light Reading1 As consumer and businesses increasingly merge user experience across multiple devices and networks, mobile networks must also be engineered to minimize delay. Calculating Total Latency The total latency budget is measured in milliseconds (ms), but is comprised of individual network elements and interfaces, which individually may only add microseconds (亮s), but must be added together to calculate total end-to-end latency. Latency must be carefully managed and measured. Latency includes delay from propagation, buffering and queuing, transmission, and signal processing that is introduced at every link and network element through which a packet travels. A primary design objective for any individual network element (such as a security gateway) should be to minimize its latency contribution in order to stay within the overall latency budget prescribed for the link. As mobile technology has evolved, the latency targets have become increasingly stringent. The figure following shows the progression of roundtrip times from GSM/Edge to LTE, as defined by 3GPP. 1 Light Reading, LTE, A latent problem
  • 5. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 5 Figure 1. Evolution of Mobile Network Latency Targets2 As shown in Figure 1, for LTE networks, the 3GPP target latency budget for the user plane latency (radio and core) is about 20 ms, excluding the backhaul transport network. The backhaul network (RAN edge- Core edge, including security gateway) would add an additional 10 ms.3 Stringent New Requirements with LTE-A In order to support the predicted growth, the 3GPP has developed LTE-Advanced Release 10, a major enhancement of the LTE standard deployed in many mobile networks today. The new technology targets peak data rates up to 1 Gbps and introduces new RAN concepts with the ultimate goal of designing a system that is drastically enhanced in both cell capacity and coverage. In LTE Advanced, latency targets are reduced substantially to 10 ms, allocated as shown following. Figure 2. Target Latency Budget for LTE-A4 2 Nokia Siemens Networks, LTE-capable transport: A quality user experience demands an end-to-end approach 3 Qualcomm, Latency in HSPA Data Networks, July 2013.
  • 6. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 6 Microsecond Performance for X2 To maintain performance requirements in LTE-A, the latency on the X2 links are especially critical: In order to meet stringent latency requirements of less than 1 millisecond, the physical and logical path of the X2 interface needs to be as short as possible. 5 In LTE-Advanced, new RAN optimization techniques introduced in LTE-A impose critical performance demands on the X2 interface, requiring very short latencies of <1 millisecond across the backhaul network. This is further described: Inter-cell interference coordination (ICIC) and coordinated multi-point (CoMP) transmission are two techniques defined in LTE-Advanced specifications that target a better user experience at the cell edge. ICIC is limiting cross-talk by coordinating spectrum allocation across multiple cells. CoMP allows multiple base stations to simultaneously serve a user device and increase the receive power level and, therefore, capacity. Both synchronization techniques are implemented over the X2 interface and require very short latencies of <1 millisecond across the backhaul network to achieve real-time coordination between base stations. In addition to providing low-latency connectivity, base station clocks need to be in phase to enable proper operation of ICIC and CoMP. This leads to the requirement for highly accurate phase or time-of-day synchronization. Most 3GPP base station clocks are currently synchronized on frequency only, since accurate phase synchronization was not a requirement up to now. The new LTE-Advanced functions, however, require base stations to be in phase with an accuracy of 500 nanoseconds to efficiently operate ICIC and CoMP. This is nearly impossible to achieve without active time distribution over the X2 interface. Mobile operators will also introduce LTE TDD radio interfaces operating in unpaired spectrum. Many operators already have acquired unpaired spectrum, since TDD provides more flexible scaling of the up- and down-link capacity and has additional benefits to the overall architecture of the radio access network. Like ICIC and CoMP, LTE TDD also requires phase alignment of neighboring base stations over the X2 interface in addition to the traditional frequency synchronization used in mobile networks today. 4 LOLA, Presentation of WP2 Scenarios and Target System Architectures 5 RCR Wireless Readers Forum, New Backhaul Challenges are emerging with LTE Advanced, July 15, 2013.
  • 7. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 7 These techniques mandate tighter coordination between base stations and places specific performance requirements for the X2 interface with respect to capacity, latency and synchronization. Supporting these requirements is not trivial since most mobile backhaul architecture on which X2 interface is transported follows a hub-and-spoke design where traffic distribution and re-direction is performed at the security gateway in the distribution layer of the backhaul network. X2 Delay and User Throughput Latency on the X2 interface reduces the benefit of co-ordination schemes. In the figure following, X2 latency as low as 5ms reduced cell edge and median user throughputs by over 20% in this example of a Joint Transmission CoMP scheme.6 This loss of throughput impairs the user experience and reduces network efficiency. Figure 3. Impact of X2 delay on user throughput with CoMP scheme 7 Simpler schemes may not be impacted as much, but this shows that the sophisticated CoMP schemes require very low X2 latencies to achieve their full potential and delivering the anticipated enhancements in speed, coverage and overall quality of experience to the end subscriber. The table below uses the 5ms and 1ms figures above and calculates the throughput loss for 200 亮s. X2 Delay 5 ms 1 ms 200 亮s Throughput Loss (20%) (5%) (1.0%) Figure 4. Loss of Spectral Efficiency from X2 Delay. 6 Qualcomm, Backhaul Requirements for Centralized and Distributed Cooperation Techniques, 8 July 2010. 7 Centralized Scheduling for Joint-Transmission Coordinated Multi-Point in LTE-Advanced, S. Brueck, L. Zhao, J. Giese, M. A. Awais, proc. ITG/IEE Workshop on Smart Antennas, Bremen (WSA'10), Germany, February 2010
  • 8. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 8 Latency Impacts the User Experience High latency causes noticeable delays in, for example, downloading Web pages or when using latency sensitive applications such as interactive games, VoLTE, M2M, and streaming media.8 Lower Latency Improves Page Load Times Analysis by Google shows a relationship between page load time and round trip latency. Their analysis shows reductions of every 20 milliseconds in round trip latency reduces page load time by 7-15%.9 The figures following compare the impact on page load time from two different performance characteristics - bandwidth improvements and round-trip-time (RTT). There are diminishing returns as the bandwidth gets higher, but not for improvements in round trip time (latency). Google concluded that, unlike improvements to bandwidth, reducing the round trip time always helps the overall page load time. Figure 5. Comparison of impact on page load times RTT and Bandwidth The linear relationship shown in the charts provided by Google suggest that microsecond latency improvements would also improve the page load times and therefor have some impact on the user experience. Both Amazon and Google have confirmed that in e-commerce applications, milliseconds of higher page load times, which is not consciously perceivable to the user, can still have an impact on user behavior and experience and can directly impact revenue. 8 Nokia Siemens, Latency The impact of latency on application performance 9 Google Blog, More Bandwidth Doesnt Matter (much ) , April 2010
  • 9. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 9 Any increase in the time it takes Google to return a search result causes the number of search queries to fall. Even very small differences in results speed impact query volume.10 Experiments at Amazon have revealed similar results: every 100 ms increase in load time of Amazon.com decreased sales by one percent.11 Estimated Sales Impact12 Lower latency, even in microseconds, improves page load times, which in turn can positively impact revenue for mobile operators and enterprise customers. To illustrate the potential impact on sales from less efficient network nodes, the table below shows the potential impact of additional 200 亮s latency (page load times) on an on-line enterprise with $14.8B in annual sales, using the 100 ms/1% sales decrease provided by Amazon. As shown in Figure 6, the estimated impact on sales for each additional 200 亮s of page load delay could be as high as $300k annually. Example - Sales Lost Per 100 ms Per ms 200 亮s 2007 Annual Sales : $14,800 M Negative Impact on Sales (1.0%) (.01%) (0.002%) Sales Lost ($148 M) ($1.48 M) ($0.3 M) Figure 6. Estimated sales lost for 200 microseconds delay. M2M and On-Line Gaming Two emerging, massive applications also require low latency: High Performance On-line Gaming M2M, Sensory applications Ericsson expects that in 2020 there will be 50 billion devices connected and available to be used in various existing and new applications. The figures below compare the latency requirements of new and existing applications and provide latency targets of a few representative applications. 10 CNET News.com/ZDNET.com: Google卒s Marissa Mayer: Speed wins by Dan Farber, November 9, 2006. 11 Nokia Siemens Networks: Latency: The impact of latency on application performance, 2009. 12 In order to estimate the impact of an additional 200亮 of latency, a linear relationship with the 100ms-level data available is assumed.
  • 10. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 10 Figure 7. New High Growth Applications require Low Latency13 M2M and Gaming Application Latency Target On-Line Gaming 80 ms (U-Plane, 1 way) Gaming Sports Events 25 ms (U-Plane) Sensor-Based Alarms 2-12 ms (1 way, U+C plane) Figure 8. Example latency targets of M2M and Gaming applications.14 Implications for Security Gateway The necessity to meet the <1 millisecond budget of the X2 interface in turn places strict requirement to minimize IPsec and packet routing latency in the security gateway, which is the key nodal element responsible for security and routing of control and data plane traffic in X2 interface. The entire S1 interface budget is only 1 ms (1,000 亮s) each way). Therefore, the example node referenced earlier that contributed 250 亮s of latency, when placed on the S1 would be consuming 25% of the entire backhaul latency budget. Each additional 200 亮s delay from a security gateway would cause a loss of 1% throughput. In contrast, the STOKE速 Security eXchange is engineered to contribute only 30 亮s or 3% of the 1 ms latency budget. 13 Eurocom, Achieving LOw-Latency in Wireless Communications 14 Eurocom: Presentation of WP2 Scenarios and Target System Architectures
  • 11. Latency Considerations in LTE STOKE速, and the Stoke logo are registered trademarks of Stoke, Inc. Copyright 息2014 Stoke, Inc. All rights reserved. Literature # 130-0029-001. 11 Conclusions Network latency, even more than download speeds, directly impacts the user experience and bottom line revenue for on-line businesses. In high frequency financial market trading, microsecond improvements are considered a competitive advantage. As low latency applications grow in importance and e-commerce increasing moves to mobile platforms, operators will need to carefully manage and measure latency budgets to maintain their own competitive advantage. The latency contribution of all individual network elements, including the security gateway, must be carefully calculated. In LTE-A and especially for the X2 interface where the latency targets are drastically reduced, an additional 200 亮s delay is a significant difference. Stoke速 Security eXchange Operators are challenged to integrate networks and technologies smoothly, sustain a quality user experience with high network performance, and still keep service delivery costs low. Stoke solutions are purpose built from the ground up for today's mobile broadband environment to solve critical, performance impacting problems for mobile network operators. Stoke innovative design and patent pending technologies enable cost effective, concurrent operation of critical functions while maintaining line-rate, high performance throughput. STOKE Security eXchange is a carrier grade, field proven solution ideal for these requirements since it introduces < 30 microseconds of latency, supports creation of multiple VLANs and line rate/10Gbps throughput for small packet sizes of 96 bytes.