Advanced Congestion Control Technical Briefing - TCP RNA
5/287 01-FGB 101 623 Rev A
Advanced Congestion Control
Copyright © Ericsson AB 2014. All rights reserved. No part of this document may be reproduced in any form without the written permission of the copyright owner. Disclaimer The contents of this document are subject to revision without notice due to continued progress in methodology, design and manufacturing. Ericsson shall have no liability for any error or damage of any kind resulting from the use of this document. Trademark List All trademarks are properties of their respective owners. A list of trademarks is found in document Trademark Information, 006 51-CSH 150 090.
5/287 01-FGB 101 623 Rev A
2014-07-10
Ericsson AB 2014 Commercial in confidence
2 (7)
Advanced Congestion Control
1
Executive Summary Two main application types are especially interesting for TCP (Transmission Control Protocol) optimization, web browsing and video. Web browsing benefits from TCP optimization as it is very latency sensitive. Also, web browsing is the main application of most people’s everyday Internet experience. Video is important as it is the most significant application in terms of volume. According to the June 2014 Ericsson Mobility Report, mobile data traffic in Q1 2014 exceeded total mobile data traffic in 2011. Data traffic grew by 65 percent between Q1 2013 and Q1 2014. The rising number of smartphone subscriptions is the main driver for mobile data traffic growth while users consume ever more data per subscription – mainly driven by video. Mobile video traffic is expected to grow by 13 times between 2013 and 2019. To meet this explosive growth in traffic, and ever increasing consumer expectation of high quality delivery of content, operators must make use of their networks in the most intelligent and efficient ways possible. One of the many ways Ericsson is contributing to this challenge is our unique algorithm for advanced congestion detection, called TCP Radio Network Aware (RNA). Using our patented algorithm in the Multiservice Proxy (MSP), based on multiple variables including round-trip time, bit rate, and packet-loss, Ericsson is able detect network congestion. This can be achieved with much greater granularity than traditional methods, which are typically based on packet-loss alone. Combining this advanced congestion-awareness with other MSP logic, such as subscriber plan, polices, user profile, and content type, will enable operators to provide a better quality of experience, like faster page loading and file transfers to subscribers than ever before. All while driving efficiency in the network.
2
The Challenge Network operators face an increasing challenge in continuously providing a high quality of experience for their consumers, while maintaining corresponding network investments and hence sufficient margins to satisfy shareholders. There is a struggle to keep the network costs (CAPEX/OPEX) down, to push up the net revenues, and to avoid churn. As the Operator’s network is continuously benchmarked towards other networks in the country, it is very important for the Operator to provide a competitive network performance and differentiate the service offering for different user categories.
5/287 01-FGB 101 623 Rev A
2014-07-10
Ericsson AB 2014 Commercial in confidence
3 (7)
Advanced Congestion Control
Mobile broadband 3G and 4G LTE networks reacts similarly in terms of throughput reductions and increasing delays caused by increasing load or reduced signal strength. Operators need to detect congestion and reduced signal strength and take actions to avoid impact on network and consumers quality of experience. Standard TCP detects congestion from packet loss only. In a radio access network (RAN) with very low packet loss, this behavior leads to excessive filling of buffers. The sender has no knowledge of the actual user equipment (UE) receive rate and thus continues to increase its send rate. This leads to wasted network resources and increased latency. Also, many browsers open up several TCP connections in parallel, each with an initial send rate. The browsers try to push hard to give each flow an advantage over other flows. This can cause the opposite effect of excessive congestion when large buffers are filled.
3
The solution
3.1
TCP Congestion Control TCP is the protocol workhorse on the Internet. Web browsers use TCP when they connect and request web pages and data from origin servers on the Internet. An important aspect of TCP is the congestion control. A TCP congestion control algorithm (CCA) is used to avoid congestion situations and at the same time give as high throughput as possible. To achieve improved QoE with limited additional RAN and Core Network investment, Ericsson has developed a radio-network-aware CCA that is used in the Ericsson Multiservice Proxy (MSP). MSP is an advanced HTTP- and TCP-proxy typically located in the GGSN/PGW premises. One TCP connection is between the user equipment (UE) and MSP, and one is between MSP and the origin server. TCP RNA is a delay based CCA and takes advantage of the proxy being in control of all TCP flows to the UE.
Figure 1 MSP and TCP RNA in the Traffic Flow
5/287 01-FGB 101 623 Rev A
2014-07-10
Ericsson AB 2014 Commercial in confidence
4 (7)
Advanced Congestion Control
3.2
Ericsson’s TCP Radio Network Aware (RNA) TCP RNA is a hybrid CCA that is integrated in MSP. It detects congestion and reduced signal strength based on both latency and packet loss. Using TCP RNA removes many of the drawbacks of standard CCAs such as no delay detection, high sensitivity to packet loss and excessive filling of buffers. Using TCP RNA, operators are able to: 1.
Manage the available capacity of their radio network down to flow granularity. TCP RNA detects an increasing congestion manifested as an increased delay before packet loss occurs. The increased delay is also an indication of the queue buildup within the network.
2.
Link the knowledge of the bandwidth availability and bandwidth management with the application and service plane for optimizing and prioritizing, contextually to the network conditions, each application for every customer.
When the latency increases, TCP RNA progressively reduces the send rate. It can also give higher priority to specific content types, protocols or subscriber categories. As an example a video stream can be given priority over a competing application update download when the delay suddenly increases.
3.3
Functionality TCP RNA is an Ericsson-developed radio-network aware Content Congestion Algorithm. It is implemented in MSP as a TCP kernel module, highly configurable from the MSP proxy layer. TCP RNA provides the following functionality:
5/287 01-FGB 101 623 Rev A
•
Real-time prioritization of TCP flows – Possibility to differentiate TCP settings based on many criteria, for example subscriber category and accessed application
•
Policy enforcement – Policies with different goals, for example faster startup of streaming video or better QoE for gold subscribers, can be enforced using Subscriber Plans and application specific business logic.
•
Smart subscriber prioritization – TCP flows generated by Gold subscribers are prioritized only when needed. TCP RNA allows prioritization of latency sensitive applications (for example streaming video and web browsing) while non-latency-sensitive applications (for example file downloads) are not prioritized.
•
Congestion aware proxy services – congestion level per device is calculated using advanced equations. The congestion information can be combined with MSISDN and detected congested areas to apply differentiated content optimization techniques to reduce the load.
2014-07-10
Ericsson AB 2014 Commercial in confidence
5 (7)
Advanced Congestion Control
•
3.4
TCP flows can be optimized also when data is encrypted, e.g. HTTPS.
Key Benefits The key benefits with TCP RNA are:
3.5
•
Application-aware – Faster time to latency sensitive content. Consumers will maintain and get even greater quality of experience when the operator can give higher priority to those applications that are sensitive to latency e.g. voice, chat, web browsing, video streaming and Facebook.
•
Reduced TCP retransmission – TCP RNA detects congestion and reduce send rate before packet loss starts to occur
•
Protects the network from bufferbloat – TCP RNA ensures that the latency stays within limits during large file downloads
•
Decrease video stalling – possible to use larger initial send rate for video content
•
Considerable throughput improvement compared to standard TCP
Live deployments and true operator results TCP RNA is deployed in several live networks worldwide. Recent engagements in Turkey and Australia have shown excellent results: 43% reduced download time for webpages – webpage acceleration (statistically proven with 100+ iterations per page) and up to 900% acceleration of downloads in packet loss scenarios and 70% faster file transfers
4
Conclusion
4.1
Why Ericsson TCP RNA has been developed based on more than ten years of research within Ericsson. It is custom made for radio access networks and improves quality of experience and reduces problem of excess buffering. The deployment of TCP RNA is within Ericsson’s Multiservice Proxy, a market leading proven, robust and advanced product with more than 140 installations worldwide.
5/287 01-FGB 101 623 Rev A
2014-07-10
Ericsson AB 2014 Commercial in confidence
6 (7)
Advanced Congestion Control
5
Contact Ericsson To understand more about how Ericsson’s Multiservice Proxy, the TCP RNA and its broader portfolio in optimizing networks for media and video delivery please contact your local Ericsson office http://www.ericsson.com/ourportfolio/tv_sales_and_support_contacts
5/287 01-FGB 101 623 Rev A
2014-07-10
Ericsson AB 2014 Commercial in confidence
7 (7)
