| Domestic mobile Internet market is booming with the widespread use of intelligent terminals and the popularity of OTT (Over the Top) applications such as social network or instant message. Does not match with the innovation of network technology and growth of network traffic and expansion of network capacity, there are declines in the user’s ARPU and the quality of the network, telecom operators are facing the risk of becoming a traffic pipe in the age of mobile internet. These years, various carriers have strongly recognized the necessity for deep understanding of the mobile packet service traffic, especially for mining basic traffic rules from the existing network real data, to enhance the utilization of existing network resources and make the "pipe" intelligent. Nowadays, domestic operators’mobile packet networks are under the deployment and expansion of3G network, for the existing network capacity and users, GPRS and EGPRS network will still on service for a long time.In this context, base to huge amounts of real data collected by passive measurement technology from one domestic operator’s on service GPRS/EGPRS network, this paper researches on the basic issues of mobile packet traffic distribution, dedicated to provide easy-operated methods to improve the wireless resource consumption and network performance for domestic mobile telecom operators in current situation.The main research contents and innovations are as follows:(1) Based on IP-based measurement, this paper puts forward an universal software architecture for mobile packet service data acquisition systemTelecom operators have to do extensive testing and analysis based to the real traffic data from their network. Due to the evolution of network technology and coexist of2/3G and LTE network, there are so many interfaces and communication protocol stacks need to be measured in communication network. These result to big challenge for rapid deployment of measurement software and easily maintenance. Benefit from the trend that the whole network equipments are becoming to support IP protocol, this paper innovatively puts forward an universal-model based measuring software framework. The framework has three main features:Plug-in for data packet parsing:Plug-in for one protocol only needed to implement once, reused in different protocol stacks. Set different connections between plug-ins to fit analysis path of different protocol stacks.Universal processing module for matching signals in control plane: By summary the features of Gb interface main signal processes in control plane, abstract general fields (such as mobile cell, user, time, direction) and featured fields of different protocol. A few matching logic can be used to adapt the vast majority of signal monitoring demand.Processing module in user plane can process date packet of user plane to a common5-tuple flow record based to the matching rules, and correspond universal maintenance field to signal record in control plane to meet the needs of real-time query.(2) Association algorithm of mobile packet traffic distribution and wireless resources occupationTraditional research and analysis on the distribution of mobile packet traffic and resource consumption is fragmented or incomplete. Thanks to the research team’s advantage at data measurement and collection, this paper based to nearly2800GPRS/EGPRS cells’Gb interface packet data collected during in continuing30hours at urban centers of one city in China in2012, analyzed the composition of protocols in current mobile packet traffic, divides the traffic into three parts(the control plane overhead, protocol overhead and user pay load), proposes a set of analysis algorithms to find out wireless resource consumption of each part of the traffic in different encoding scheme. The study finds out:In terms of the composition of traffic:user plane traffic accounted for more than99%of the overall traffic in wireless interface, and95%of these traffic using the TCP protocol, and3%of the traffic using the UDP protocol; control plane traffic accounted for less than1%of the overall traffic. In terms of the wireless resource consumption:for PDTCH resources, the overhead of TCP and UDP protocols may account for30%of the total consumption of resources; the overhead of GPRS control plane can be negligible. Protocol overhead consists of cumulative message heads, TCP control messages and retransmission messages. Large protocol overhead is mainly due to user plane packet payload is generally small and the high proportion of pure TCP control packets. For CCCH resources, this paper innovatively proposes an algorithm to restore TBF and determine the scenarios for TBF establishment, and analyzes the impact of the variable algorithm parameters, and verifies the efficiency of the algorithm. This study finds out that small flow TBF accounts for about40%of the total TBF establishment number. Small flow TBF means that in a single transmission the flow isn’t more than100bytes. At the same time, heartbeat TBF accounts for about70%of the small flow TBF, paging triggered by heartbeat TBF accounts nearly three times to non-heartbeat TBF. Experiments show that when threshold of TBF release delay is increased from1s to10s, the number of concurrent TBF doesn’t significantly improve, but CCCH resources can save20%. This shows that when the CCCH channel resource constraints, besides of channel expansion and heartbeat rate adjusted by service providers, telecom operators can adjust the threshold of TBF release delay to relieve the channel pressure.(3) Mobile packet traffic cell-time distribution modelDerived by analyzing the real data traffic volumes of mobile packet core network per cell in5minutes during whole day, this paper finds out data packets arriving volumes per5minutes of any mobile cell has a better fit with exponential distribution. The result shows that most of cells at most time are at low speed state; and high burst traffic is mainly concentrated in a small number of cells in a few hours. This study also uses other data samples collected in different regions and at different acquisition time to verify that arriving traffic volumes obey exponentially distribution has certain universality. The research can help engineers and researchers to understand the current network GPRS/EDGE cell flow performance throughout the day by simple parameter estimation. (4) Mobile packet service network capacity performance analysisBased to5minutes real traffic data measured on the current cell, analyzes GPRS network packet loss-ratio and influencing factors, then this study finds out that LLC-discarded discard ratio has a positive correlation with concurrent number of users in a cell, and has little correlation with the volume of traffic of5minutes in a cell. The research shows that bottleneck of the current cell capacity is due to limited number of concurrent users, not due to limited channel bandwidth.Under the situation that the majority of traffic is a small flow, telecom operators should increase the upper limit of concurrent TBF appropriately to reduce discard ratio. The study also shows that the discarded ratio of cell reselection has little correlation with concurrent number of users in a cell and the volume of traffic of5minutes in a cell. This discovery means that cell reselection hasn’t been used for a method to balance cell capacity in GPRS/EGPRS system. As cell reselection discarded has nothing to do with network capacity congestion, analysis on GPRS network congestion can mainly focus on monitoring LLC-discarded discard ratio. |
PDF Full Text Request