Rx for Cell Congestion
Mobile networks continue to be particularly prone to delayed data sessions: the web page that won’t load, the flat grey Instagram “picture” -- with caption, the spinning wheel of time over a video player. Experiences like this can be especially frustrating because these challenges have substantially been solved on other types of networks through quality-of-service (QoS) approaches.
A new era of mobile network technologies that meaningfully improve user experiences is emerging. Dynamic rate control with feedback (DRCF) has demonstrated the promise of better mobile performance positively affecting customer behavior. Rate control is the practice of modifying data flows to manage demands on resources and capacity. The dynamic application of it is needed to adapt to the ever-changing demand and capacity, such that control is applied only when congestion is impacting user experience. And feedback from actual real-time performance is crucial to comprehend and best respond to that constant change.
The Cell’s Congested Legacy
Wireline networks have experienced congestion throughout their existences. As traffic grew in the 1990s and data services became ever more important, various approaches were developed and refined to improve QoS. Traffic Management (TM) schemes were implemented in routers, switches, cable modem systems, and DSL multiplexers to handle the situations of too much traffic trying to squeeze through limited-size pipes. One of the schemes to manage bursty traffic into fixed capacity links has been “leaky bucket” rate-shaping algorithms, in which each session at its initiation is paced to get its turn for resource allocation.
These schemes are generally augmented with weighted fair queuing (WFQ) mechanisms, in which traffic classes like best effort or expedited forwarding are queued and weights are assigned to the queues. The fixed capacity is then allocated to the classes based on the weights. If an important class doesn’t use its bandwidth for a few moments, then that bandwidth can be given to a lower priority class that needs it. This technology is how a cable operator can offer a voice service, or a DSL provider can offer a video service.
While these technologies are very much proven to deal with bursty demand on fixed size pipes, they fail to deliver quality in a mobile environment. If applied to these environments, they would result in both frustrated demand and inefficient use of expensive cellular infrastructures. There are a number of factors that render these legacy traffic management technologies ineffective:
1. Fluctuating bandwidth. The bandwidth that is shared by the users in a cell is not fixed, and fluctuates moment-to-moment as the users move around within the cell and switch between cells. The pace at which the bandwidth fluctuates is fast enough to significantly interfere with TCP’s mechanisms.
2. Unknown bandwidth. The actual shared cell bandwidth is not known until the bits are pushed out the antenna and make it expeditiously to the user equipment. As a user passes behind a building or into an elevator, it can instantly affect how much bandwidth is available. Until the cell struggles to get those bits to the user, it is not known how many bits per second can be achieved at that moment.
3. Uneven tradeoffs. The cell bandwidth cannot be evenly traded off among the services in a cell. In a fixed network, if one wanted to shift 50 bytes of capacity from 1 service, then the 50 bytes could be used by another service. In a mobile network, each user device may have different signal conditions. Shifting 50 bytes from 1 service could allow just 25 bytes or as many as 100 to be delivered for another service.
4. Inconsistent members. In a fixed broadband network, the number and identity of the users that are sharing the capacity of a link stay the same. The same set of DSL customers always share the same 1 gigabit link to the DSL multiplexer. But in a mobile cell, the number of users and the identity of users changes constantly. In fact, it can change in the middle of a user’s video streaming or browsing session. Therefore the set of users and services that are contending for resources is inconsistent.
The devices themselves lack comprehension of a cell’s context, other than performance of their own sessions, so when they perceive swift packet delivery they tend to request higher-speed sessions. When this occurs by multiple, adjacent devices simultaneously, it can suddenly bring on congestion within the relevant cell with corresponding spikes in latency and packet loss. Just as suddenly, experiencing the consequences of poor packet delivery, those devices can react by backing off their demands and causing a sudden decline in the cell’s resource utilization. The user experience suffers throughout.
For these reasons, legacy traffic management and rate shaping techniques do not provide the desired and expected result of high quality services. The consequences of mobile customer frustrations are increasingly clear as improving analytics are now revealing tendencies to abandon sessions after even 1 stall, which cascades with further instances. And surveys show that customers are quick to blame their mobile network operators for these problems more than providers of applications or devices. So it’s appropriate that operators respond with efforts to alleviate congestion and assure best experiences. However, legacy approaches are ineffective and accelerated upgrades are expensive, imprecise, and may not even eliminate latency and corresponding packet loss, unless some means of managing contending services is implemented.
Taking a New Direction With DRCF
Distributed networking capabilities, appliance architectures and algorithmic advances allow for highly-economical distribution of intelligent network functionality at the most crucial portions of networks. In mobile networks, the juncture between the mobile network core and the radio access network is particularly strategic because from that location, traffic is at scale for cost-effective management, yet the characteristics of each cell’s sessions can also be determined. This is the right location for applying dynamic rate control with feedback.
DRCF follows a continuous cycle, maintaining and updating the state of each cell and each session within the cell. It uses those session states and cell state to provide ongoing feedback to rate control, adapting in real-time to capacity changes and traffic demand. A primary step of DRCF is to map user equipment within each cell, including movements, entries, and exits, and which devices are active and consuming services. This step is based on 3GPP standard messaging. It is a highly-distinctive exercise compared to fixed network considerations that are far more static in device distribution and fixed in terms of capacity.
Session characteristics can be further comprehended by the type of application, such as streaming, browsing, or background updating. Operators may wish to prioritize with even greater precision, and signature libraries can be perused for more specific determination such as grouping social media applications. The industry focus is to classify categories of applications based on the parameters needed to achieve a high quality user experience; it is not to prioritize one content provider versus another. Besides classification into application types, mobile service providers are also interested in enabling a premium service class, wherein customers that pay for a higher level of service can get the most responsive service during times of congestion.
DRCF needs to take action only if there is congestion, and this is very much a live consideration, especially given the technology’s bias to not impede session flows until congestion is detected. This leaves the network prone to the same sorts of demand spikes that occur historically when multiple devices simultaneously determine to steeply increase their bandwidth requests. But with assessment of conditions at milliseconds-scale, and maintaining session and flow state, DRCF can anticipate congestion issues within a particular cell just as it occurs to know precisely when to take action.
If congestion is detected, rules are consulted to determine what actions to take. These rules can be set by the operator. In fact, at this dynamic time in mobile network growth and perhaps in part due to varying customer preferences in different regions, priorities by operator tend to vary. Some operators emphasize streaming video quality to impress with the potential of particular sessions, while others emphasize browsing to assure the best experiences for the largest number of users and some prefer to alter the policies to provide the strongest social networking performance.
DRCF in Practice
The mobile industry is implementing its first DRCF installations. Results are compelling with more than 25 percent improvements in latency reduction and in throughput increases of both video and browsing under congestion situations. Preliminary data from these deployments shows that customers watch further into video sessions with DRCF enabled. And industry research shows that latency reduction leads to more lucrative engagement. This indicates potential to improve both customer satisfaction, and increase revenues over time by providing more content. It also promises capabilities to do so economically, with strategic placement of appliances at appropriate network junctures, instead of accelerating brute force network upgrades. The overall result is mobile network operators applying economic technology towards better overall operations without the offsetting compromises they’ve traditionally been forced to accept.
About the Author
John Reister is Vice President of Marketing and Product Management for Vasona Networks, helping to drive the company’s collaboration with mobile operators to reduce cell congestion and deliver better subscriber experiences. He has more than 17 years of experience in networking and telecommunications. For more information, email firstname.lastname@example.org or visit www.vasonanetworks.com.
What is your experience with this? Tell your fellow readers now!