Abstract--- Long Term Evolution (LTE) is a significant project of 3rd Generation Partnership Project also call as a 3GPP. First proposed on the canada conference of 3GPP in year 2004 and officially started as LTE work item in year 2006. The transition from the 3rd generation to the 4th generation has been achieved great capacity and high speed of the congestion control in mobility network. It can defined a new packet wideband radio with the flat architecture and assumes a full Internet Protocol network architecture in order to assure voice supported in packet domain in design. Beside that, it also combined with top of the line radio techniques in order to gain better performance than Code Division Access approaches. LTE can provide scalable carrier bandwidths from 1.4 MHz to 20 MHz and frequency division duplexing as well as time division duplexing . In other word, The LTE technology separately in different aspects of LTE architecture and technical principles to clarify how LTE as a radio technology achieves a high performance for cellular mobile communication system. As we knows, if increase number of user, it will limits the transmission of data rate and will affect the performance of the tranmission system in the congestion control of LTE. So, we need to overcome this to find a traffic between the two eNodeBs and analyze the difference type of application between them and check for traffic congestion control. In this project, we will use the simulation NS-3 to evaluate the performance of congestion control mechanisms in LTE wireless comunication network.
Index Terms--- LTE, Congestion Control, 3GPP
Evolution of the mobile technology, 4G Long Term Evolution (LTE) is a wireless communication or mobile cellular technology that aims to provide the high speed data for users. It is a standard for smooth and efficient transmission to increase the capability and speed of wireless data network. It used the GSM/EDGE and UMTS/HSPA network technology. LTE provides downlink peak data rate of 100Mbits and uplink peak data rate of 50Mbits. The differences between LTE and 3G is the speed and frequencies. LTE quicker than 3G which mean better in QoS and QoE. It using scalable channel bandwidths of 5-20 MHz, optionally up to 40 MHz. Figure 1 show the comparison between LTE and 3G. Otherwise, compared with the differences of LTE and WiMax. One of the differences is their function on differences frequencies, making their deployment slightly differences. WiMax combines the cellular range with the Wi-Fi speed and it plugs into Ethernet like Wi-Fi and provides roaming voice and data like cellular. LTE support Multi-RAT (Radio Access Technology) with resources controlled from the network. Beside that, the LTE transmission data rate is still have a bottleneck by using high speed wired part. Without the efficient Congestion Control mechanisms, the packets will be buffered and discarded due to buffer overflow in the eNodeB.
Figure 1 : Comparison between 3G and 4G
LTE is composed of many new technologies compared with the previous generation of cellular systems. These new technologies are used to generate more efficiency with regards to spectrum and higher rates as expected by designers. Here are only snapshots of the technologies and they data will be clarified in detail in the third section. The first of the LTE technology is OFDM (Orthogonal Frequency Division Multiplex). In order to gain high data bandwidth when transmitting packets, LTE integrates OFDM technology which can provide high-degree resilience to reflections and interference at the same time. Furthermore, the access schemes can be divided into two access approaches used in the DL and UL respectively. One for the DL is OFDMA (Orthogonal Frequency Division Multiplex Access) and another one for the UL is SC-FDMA (Single Carrier- Frequency Division Multiplex Access), which has the advantages of smaller peak to average power ratio and more constant power able to get high RF power amplifier efficiency in the mobile handsets. Beside that, the second of the LTE technology is MIMO (Multiple Input Multiple Output. MIMO operations include spatial multiplexing as well as pre-coding and transmit diversity. These operations addressed the problems of multiple signals arising from many reflections, which were encountered by previous telecommunications systems. Moreover, using MIMO also increases the throughput via the additional signal paths after those operations. MIMO requires two or more different antennas with different data streams to distinguish the different paths.
Radio Access Network is required to provide the services such as file transfer, video streaming, online gaming, and VoIP with guaranteed quality of service. RAN support many users as possible in order to maximize operating revenues. The numbers of users supported with given QoS is constrained by limits on radio resources such as transmission data of bandwidth. Basically, the more resources a bearer services are allocated, then the higher QoS are achieve in the network. Congestion Control mechanism is needed to reduce the cell load in a controlled.
Modern mobile traffic is harder to control. With 4G LTE networks, packet loss does not necessarily mean there is congestion on the networks, and vice-versa, congestion does not necessarily mean that there is packet loss. This realization means that congestion control algorithms cannot focus on packet loss and latency alone for determining congestion. Performance analysis of LTE by using metric include latency, jitter and packet loss. Latency is defined as the time it takes for source to send packet of data to receiver. Latency typically measured in miliseconds. Therefore, the lower latency is the better network performance. Packet jitter is defined as the variability in packet delay within a given media stream. The packet loss ratio is the number of corrupted, drop and delayed packet.
Congestion control uses a number of mechanisms to achieve high performance and avoid congestion collapse, where network performance can fall by several orders of magnitude. These mechanisms control the rate of data entering the network, keeping the data flow below a rate that would trigger collapse, resulting in a poor end-user experience. Congestion Control is used to reduce the load when cell overload is detected. When the cell load remains above a threshold, the congestion will remove a subnet until the load is reduced to an acceptable level. Congestion control can categorized as non pre-emptive and pre-emptive. Furthermore, Admission Control is activated before the new request has been accepted else the Congestion Control is activated after the new request has been accepted.
To use the Congestion Control mechanisms in cross-traffic assistance service, because the Congestion Control mechanisms can accurately keep high priority users. In other words, it is important for safety applications. Therefore, the congestion control is suggest to use in the network. Otherwise, the information of the status of vehicles may not be known before the new request is accepted by using Admission Control may mistakenly reject high priority users. When the Congestion Control Module receive load reduction request from resource scheduling module, it can calculates priority by sorting all users based on certain criterion.
2 RELATED WORK AND BACKGROUND STUDY
Congestion Control in the LTE wireless network is divided into Congestion Control at the end system and Congestion Control at the network side. Congestion Control at the end system mainly rely on various versions of TCP protocol. While Congestion Control at the network side is achieved by carrying on active queue management in buffer queue of router, dropping packets predictably before the formation of a full queue, avoiding deadlock, full queue and global synchronization caused by packet loss. Active queue management mechanism can be divided into two categories. One based on real queue and the other based on virtual queue. Active queue management mechanism based on real queue contains Reciprocating Equipment Management (REM) and Proportional-Integral (PI) control. Based on virtual queue and corresponding to real queue, queue management has a rate less than real queue. Packet drop or mark is based on whether the virtual queue is full or not. Literature studied the congestion control mechanism combined by virtual queue and real queue. Through simulation, there comes the conclusion that dealing with burst data's robustness, queuing delay and jitter, active queue management mechanism based on virtual queue has the original active queue management mechanism.
The performance of LTE to support vechicular safety application is one of the important issues in network. Normally, it have two approaches can be used for cell load management such as admission control and congestion control. Admission control is used to block user connections when the cell load reaches the limit. Otherwise, congestion control is used to gracefully reduce the cell load either by sacrificing the quality of connections or by removing existing bearers. In other words, LTE admission control and congestion control have several mechanisms proposed. The mechanism not only takes into the frequency resource utilization but also the throughput. It provides the flexibility to the operators for managing cell load and user performance. The purpose of the pre-emption congestion control algorithm allows high priority requesting bearers to displace low priority connected bearers in order to reduce cell load. The priority based admission control can achieve low dropping and blocking probabilities by combine with the algorithm.
There exists an extensive body of literature on rate and congestion control in the Internet. However, few addresses the unique issue of bandwidth management for media delivery on mobile broadband networks. There are two major categories of congestion control and rate control algorithms in the literature:
i - Those based on the estimation of available bandwidth.
ii - Those based on end-to-end congestion control.
An example of rate control based on available bandwidth estimation has been shown that this category of technologies frequently fail in complex networks and will have issues in an environment such as mobile broadband where both loss and delay measurements have significant noise. Moreover, such approaches fail to share bandwidth fairly and thus we do not consider them. End-to-end congestion control, such as the standard Additive-Increase Multiplicative-Decrease (AIMD) algorithm used in TCP, dominates the Internet.
There have also been several attempts to improve the performance of congestion control protocols over networks with wireless links, where not all loss is caused by congestion. A good summary of existing work is provided in. However, existing work has mostly focused on loss-based protocols, with the goal being to either hide loss from the upper layer (such as TCP) via retransmissions at the lower layer, or modify the upper layer protocol to determine whether observed loss is actually due to congestion. The former approach requires changes in the hardware or firmware which is difficult to implement. As an example of the latter approach, TCP Westwood attempts to determine which loss is due to congestion by utilizing bandwidth estimation techniques to set the slow start threshold and initial congestion window. Although this works well when only loss is a noisy congestion signal, it does not work well on mobile broadband networks where queuing delay is also a noisy signal since bandwidth estimation techniques themselves do not work well.
Step 1 : Study different concept of LTE congestion control (archicture and components)
In this paper, the first step is we need to analyze the Long Term Evolution (LTE) congestion control for the current emphasis by wireless network operators that a standard of 3GPP and will be implement for the future 4G networks. Beside that, we also need to clear about the components of the LTE Network Archicture. The goal of the System Architecture Evolution (SAE) effort in 3GPP is to develop a framework for the evolution and migration of current systems to a system which supports high data rates and low latency. All IP network packet-optimized and provides service continuity across heterogeneous access networks. LTE wireless broadband in addition to data for the design of optimized performance characteristics can be with another GSM service provider's network compatible, so no matter service providers are already deploying UMTS technology may apply to the architecture of LTE deployment, which can allow service providers to provide better through the application of LTE services. The 3GPP on LTE in January 2008 had be included in the official 3GPP R8 standard, released in December LTE R8 version of FDD-LTE standard that defines Standardized by 3GPP LTE architecture also known as Evolved UTRAN architecture (E-UTRAN) where the system using the Node B (eNB) and the receiving network gateway (eGW) composed of two functional modules shown in Figure 2 and interface points of LTE architecture in Table 1. Functional modules using WCDMA Node B network transport protocol, which can effectively reduce network complexity and cost of the delay time to reach the requirements, and has a radio resource management (RRM) , User Equipment and Network QoS agreement between the location search, mobility management, and conversion between different receiver technology, security, encryption, file compression, request retransmission (ARQ), IP address assignment, multimedia broadcast and multicast (MBMS) and other functions. Strategies used between Node B transmissions grid (Mesh) way links, as defined by 3GPP next generation network architecture. 
Figure 2 : LTE Network Archicture
Table 1 : LTE/EPS Standard Reference Point 
Based on the difference between each of the system archicture regarding the eNodeB, the system architecture consists of three module shows on Figure 3 . Firstly is the Resource Scheduling Module, it allocating resources and running scheduling algorithm. Secondly is Congestion Control Module, it prioritizes the users based on certain criteria and interacts with Resource Scheduling Module to retrieve the cell load condition. Lastly is State Information Management, it stores information for current cell load and load contribution. When new connection are request, the Resource Scheduler Module will allocates available Resource Block in the resource grid. The Resource Scheduler Module sends load reduction request to the Congestion Control Module, where is the priority-based Congestion Control algorithm will be activated to reduce the system load.
Figure 3 : System Architecture of eNodeB
Step 2 : Understanding previous work related to the planning of LTE congestion control
In this step, we need to understanding previous work related to the planning of LTE congestion control. The previous work related to the planning of LTE congestion control by using congestion control could leads to temporary cell overloading. Even though the MAC scheduler could absorb the overload traffic for a short time, this could lead to increasing bit error rate and incur more retransmissions. It may not be acceptable for beacon transmissions because of the emergency information it conveys. Therefore in this paper, we assume there is a 'buffer' (i.e., reserved resource blocks) to accommodate the temporary overload. When the congestion control module receives load reduction request from resource scheduling module, it calculates priority by sorting all users based on certain criterion. For cross-traffic assistance, we propose two methods for prioritization. The first method is prioritization by distance to an intersection. Because the vehicles near an intersection are the ones that need collision warning service most, the priority of each user is calculated based on the distance of the vehicle to the intersection. The closer of the vehicle is more higher the priority. Beside that, the second method is prioritization by arrival time to an intersection. Calculating priority based on distance to an intersection may not be accurate. For example, a vehicle drives at high speed will approach an intersection more quickly than the one drives at low speed. Another way for prioritization is to calculate priority based on arrival time to an intersection. It can be calculated easily given the speed of the vehicle and the distance to the intersection. In other word, time is short, more higher the priority. Another special case is when vehicles stop at intersections waiting to cross, the highest priority to these vehicles in this situation.
Step 3: Developing a topology for the planning.
In this step, a suitable environment and a proper tool is planned and prepared for the project experimental to analyze the traffic between the two eNodeBs in 4G LTE and to capture different type of traffic and check for traffic congestion in between the eNodeBs. After discussion, we planned to use the NS-3 simulator in the Ubuntu linux environment. NS-3 is a discrete-event network simulator for Internet systems. It's goal is to create an open simulation environment for networking research that will be preferred inside the research community. C++ programming language will be used in the ns-3 simulator to produce the output of the topology. The topology will include the links and nodes, such as the UE and eNodeBs, so that we can analyze the traffic and check for traffic congestion in between the eNodeBs.
Step 4: Implementation on the NS-3 Simulation.
Since we are evaluating the performance of congestion control by simulations, so we conduct our simulation with NS-3 network simulator with LTE-EPC extension. The base station is installed in the center of the network topology and it is connected to other UE. Low priority UE should be ignored first until they regain high priority rather than disconnecting them straightaway when bandwidth is limited. This gives us a sense of how the prioritization scheme could help in managing the bandwidth utilization. After the simulation of environment between enodeB and UE, we send the different applications between the nodes. Then, the data is captured.
Step 5: Analysis and comparison of the data if there is congestion.
The data captured will be analyzed if there is congestion. If there is no increase in the number of UE, still check for the same. Comparison of data enables us to know the impact of increased traffic on the data flow and also to determine whether there is traffic congestion or not. Hence, we will be able to study different types of applications data transfer rate and check for the existence of the probable traffic congestion with increased load in traffic. The figure 4 show the evaluation the congestion control mechanisms performance by using the waterflow methodology.
Figure 4 : Evaluate the congestion control mechanisms performance by using waterfall methodology
4 RESEARCH CHALLENGES
In this topic, there are some challenges in wireless QoS or LTE QoS. The first challenge is network congestion delay. The end to end transmission of delay increase because the packet spend many time in the queue at each hop, so it happend network congestion delay. Beside that, Some wireless network span distance that are measured in kilometers. In these network, the congestion delay can be tremendous burden to all communication. Therefore, this problem may exist to some extent in etropolitan area networks and it is a signficant issues in satellite communication.
The second challenges of LTE congestion wireless QoS is enhancement for unknown bandwidth. In order to accurately calculate the feedback, the router must know the exact bandwidth capacity in advance. The output link capacity acts as an important parameter in the control algorithm. If the router underestimates the bandwidth capacity, it will underutilize the link and waste the valuable bandwidth resource. And if the router overestimates the capacity, it will give improper feedback to senders to increase their congestion windows and may cause queue growth and even buffer overflow (congestion). But it is very hard to decide a proper value of for a wireless link in advance. One reason is that a wireless channel is shared by competing neighbour nodes and the number of nodes sharing this channel may change at any time. Another reason is that the wireless link bandwidth is affected by many changing physical conditions, such as signal strength, propagation distance, and transmitter power. For example, an 802.11 node can change its MAC-layer data rate dynamically for different physical conditions, which means the output bandwidth of this node and other neighbor nodes may also change.
Beside that, the third challenges of LTE congestion wireless QoS is loss of packet. For a sender in lossy LTE wireless environment, to differentiate two kinds of packet loss. For the non-congestion-related loss (bit error), it should maintain the current window size. Another is for congestion-related loss (buffer overflow), it should slow down to prevent congestion collapse. The assumption is that packet loss may reveal a congested router in the path and transiting to standard TCP behavior is a conservative response. However, if we are sure that all routers along the path support router-assisted congestion control, such slowdown reaction should be unnecessary for packet loss caused by bit error. For TCP, the sender needs to slow down on detecting packet loss because packet loss is the congestion signal for TCP. This is due to the design rationale of TCP congestion control. A TCP flow keeps increasing its sending rate and intentionally fills up the buffer of the bottleneck router to generate packets drops. Through this approach TCP finds the available capacity of the path. But for router-assisted approach, since congestion information has already been wrapped in the special packet header and communicated to the sender, the sender should not insist treating packet loss as congestion signal now. When the queue is too full, it will start to drop the incoming packets. However, typically suffers much more loss due to data being corrupted during transmission. This interference may come from other communication occuring on the same frequency and also electrical noise.
Lastly, one of the challenges of LTE congestion wireless QoS is uncontrolled routing can cause by uncceptable delay or jitter. Which mean the two packet are sent, normally no guarantee they will take the same path to the destination. Therefore, if one path has more hops than more congested andhave been the packet will not arrive at the destination at the same time. 
In this paper, we analysis congestion control mechanisms for LTE network, the following several cases are considering such as wireless network consisting, mixed wired and wireless network congestion. Congestion control mechanisms of Machine Type Communication network requires the congestion caused by the co-ordination of the above three reasons. The manifestation of network congestion is the excessive number of TCP connections in the network. The forwarding capacity is limited, the decline in network transmission capacity and reliability drop, then known as network congestion. When congestion is severe, the throughput may be a serious decline, it was said that network crashes. Some part of Machine Type Communication network was constituted of wireless network, there are congestion causes the following. Firstly is frequently switch impacts on network quality. The network performance will affect when the network switching as wired to the wireless switch, wireless to wired switch, and location update of mobile devices lead to network data traffic switching. Secondly is the bit error rate on the transmission performance. A wired network has a good Error Correction Mechanism, but wireless network dues to its uncertainties, resulting in a variety of unpredictable error, no effective mechanism for error handling, and can not be effective recovery. Thirdly is no effective mechanisms of error monitoring. A wireless network can only monitor the discarded packet not be effectively discarded because of an error was detected, and it cannot be correct according to the specific cause of the error recovery.
Beside that, to proposed a priority-based congestion control algorithm for cross-traffic assistance using LTE networks. We should define the service region within which the cross-traffic assistance should offer service. The priority is assigned based on the distance to an intersection or the arrival time to an intersection. The simulations show the methods can effectively control the cell load while provide service based on the urgency of users. Based on different vehicular safety applications, different criteria may be used for prioritization. For example, beacons sent by vehicles moving at high speed may be given higher priority than those sent when vehicles are at low speed on highways. In other word, beacons sent by vehicles taking lane changes should be given precedence over others. For every safety application should have a priority scheme due to the limited bandwidth and for the flexibility of load management. Furthermore, priority can also be given to different services, allowing the prioritization among services. At the end, we able to study difference type of application data transferred between the eNodeBs and the impact of it on data transfer rate. Beside that, we can analyze and check for the existence of the probable traffic congestion control with increase load in traffic.