Chapter 3 Resource Management Issues and Problem Statement
3.4 Problem statement
3.4.3 Performance metrics
In the traffic report system, its performance can be evaluated by location coverage, duplication ratio and overall traffic data value. The performance metrics to evaluate our design are shown below: one of our objectives. The definition of duplication ratio is:
Where represents the total number of received traffic data entries, and is the total number of duplicated data entries.
Overall traffic data value:
In our design, each vehicle, sensor/device and location has its own importance degree, which indicates the traffic data value that it can provide. Therefore, when
allocating resources, we have to consider the traffic data value of each sensor/device.
The definition of overall traffic data value S is:
∑
{
Where value represents the overall value of the traffic data we obtained. means the traffic data value provided by the th sensor/device if we allocate resources to the th sensor/device. is the number of sensors/devices.
Chapter 4
Related Work
The researches related to resource management problems in LTE can be roughly classified into two types: (1) non application-aware (2) application-aware mechanisms.
Figure 2 is the classification tree of related work. In addition, application-aware resource management mechanisms can be applied either to mixed-traffic, which includes various types of data, or just single-type traffic. Our work is dedicated to application-aware resource management for mixed-traffic over LTE, and the application we concerned is the traffic report system. Follows are brief review of related work:
R. Kwan et al. [4] proposed an uplink resource management algorithm based on PF Figure 2 The classification tree of related works.
(proportional fair) basis. The authors focused on the multiuser fairness problem of resource scheduling strategies, and they also provided a suboptimal solution with low complexity. By leveraging their PF scheduler, the variations of user bit rates can be effectively reduced compared to Max-Rate scheduler.
I. C. Wong et al. [3] presented a resource management algorithm for uplink LTE.
They modeled the resource allocation problem as a pure binary-integer program, and provided a greedy-based heuristic approach with low complexity. Their simulation results indicate that the spectrum efficiency can be improved by 50% using their algorithm while using the optimal algorithm can be improved by 100%.
H. Luo et al. [21] presented a cross-layer weighted-RR (round robin) QoS-aware resource management solution for downlink LTE. They focused on the application layer QoS requirements to optimize the video transmission. In their research, they jointly concern channel quality, delay-constraints and user fairness among all users against each resource block, and further dynamically adjust the MCS (modulation and coding scheme) and encoding parameters to achieve better video quality. The performance evaluation results show that the video’s PSNR is significantly improved by adopting their method.
L. Li et al. [15] dedicated to the mapping function between QoS class identifier (QCI) to Diffserv Code Point (DSCP) (QCI/DSCP mapping), which can extend the QoS provision from the bearer-level to the transport-level. The authors also presented a system level performance management tool for end-to-end QoS monitoring.
A. Gotsis et al. [2] introduced the challenges and solutions of M2M scheduling over LTE. The authors pointed out main issues when M2M comes to LTE: numerous devices, sparse transmission, and widely range of applications. These issues lead to signaling overhead in the control plane, and the diverse QoS requirements are hard to
granularity of scheduling. The idea of group-based scheduling is to schedule resources among device-groups rather than devices themselves, while time granularity of scheduling focused on the scheduling time period.
El Essaili A. et al. [1] proposed a QoE-driven optimistic resource allocation method for LTE uplink. They formulize the Quality of Experience (QoE) value, by using the mean opinion score (MOS) function to determine the quality of user experience. This research focuses on the popularity of user generated video contents transmitted over LTE radio networks. The popularity is used to rank the video contents, while the network control entity (e.g., eNodeB) can schedule resources among content producers.
S. N. K. Marwat et al. [5] proposed a scheduling algorithm based on weighted-PF on multimedia applications, called BQA (bandwidth and QoS-aware scheduler). They concern throughput, delay, fairness and the different QoS requirements of each traffic types, and allocate resources based on weighted values calculated from the above parameters. The results show that with the help of QoS weights, each traffic bearer can attain better performance.
In our research, we focus on the end-to-end resource management for the multimedia traffic reporting system over LTE, and the objective is to gather more data in a specific area. The qualitative comparison among our work and related work is shown in Table 3. The main feature of our work is that we allocate resources not only by channel availability but by information and location importance degrees.
Table 3 The qualitative comparison of existing resource management mechanisms and the proposed QoS-MTRS.
Classification Main Issue Approach
Non application-aware Uplink radio
Chapter 5
QoS-MTRS: QoS-aware Resource Management for Multimedia Traffic Report Systems
In this chapter, we introduce our proposed QoS-aware resource management for multimedia traffic report systems (QoS-MTRS) in detail. QoS-MTRS is an end-to-end resource management mechanism over LTE. It leverages the functions of EPS bearer and policy controller to support the QoS provisioning. The resource management mechanism in our design is used to guarantee the performance of multimedia traffic report systems.
To derive the most complete and valuable traffic data under certain channel conditions and limited resources, the proposed QoS-MTRS acts as a policy controller in the LTE packet core to make all the allocating decisions. Its responsibility is to decide how bandwidth resources and QoS parameters are reserved to each bearer.
The overall system flowchart of QoS-MTRS is shown in Figure 3. In our design approach, there are two important functional modules, information importance analysis and channel availability check, which are described in section 5.1 and 5.2:
5.1 Information importance analysis
In this module, we analyze information importance of all the traffic data and check for traffic data duplications. First, we have to compute the traffic data value of every on-car sensor/device. Because of limited channel resources, choosing the most worthwhile information to transmit can prevent resource wasting. There are three variables can affect of each sensor/device, which are vehicle type, sensor/device type, and location. For example, a video traffic data from a police car at a crossroad may be more urgent than an audio one from a household vehicle at an alley. Therefore, we compute of all the sensors/devices for later sorting.
Figure 3 The overall system flowchart of QoS-MTRS.
Another issue for this module is the traffic data duplication problem. For instance, vehicles at the same location may report identical traffic data at a certain time if they are equipped similar sensors/devices. Hence, checking whether there are duplicated traffic data and removing them is required in saving resources.
5.2 Channel availability check
When allocating resources, it is necessary to make sure radio channel quality and the transmission environment are suitable for sending data. The CQI must be measured and concerned to prevent transmission failures and resource wasting. For example, if a vehicle is under poor channel condition, it may be allocated more radio resources to transmit its data because the data rate has been reduced due to the poor channel. Therefore, checking CQI and derive the required resources is necessary for efficient resources management.
5.3 EPS bearer establishment and QoS parameter setting
When a vehicle enters the LTE network, it must establish an EPS bearer for IP network services. The default EPS bearer is used to support basic transmission and signaling control for dedicated EPS bearer establishment. The dedicated EPS bearer is the level of granularity for bearer-level QoS control [9], which indicates the bandwidth resource allocation in the form of multiple QoS parameters, including QCI, GBR and MBR. In our design, the resource manager acts as a policy controller to manage the resource arrangements and the QoS parameter settings.
5.4 Design approach
In our design, there are five stages in the whole process, and here we describe each stage in detail:
5.4.1 Deriving pre-defined QoS requirements
In the beginning, resource manager will derive the QoS requirements from back-end database for later usages. These requirements include importance degree and guaranteed bitrate of each vehicle type, sensor/device type, and location. The importance degrees are regarded as given values because they are derived from numerous statistics data or based on the needs of service providers. There are three types of importance degrees:
Importance degree of vehicles:
Traffic data reported by different vehicle types may have different importance degrees. For example, the importance degree of a police car is higher than that of a household vehicle.
Importance degree of sensors/devices:
The type and ability of each sensor/device can cause the divergence of reported traffic data. For instance, video data can provide better understanding of multimedia traffic conditions than text data, and the resolution of a video affects the quality of the reported data.
Importance degree of locations:
For traffic report system, some locations like crossroads need more attention for safety monitoring. We partition a map into several blocks as shown in Figure 4, and each block takes a given importance degree. Note that the map image is from the Taipei e-Map [22].
5.4.2 Vehicle connection establishment
After deriving all the needed QoS requirements, the resource manager waits for vehicle connection requests. When vehicles enter LTE network and try to establish default EPS bearers, the information of sensors/devices equipped on these vehicles will be transmitted as request payloads. These sensors/devices information will be sent to the resource manager for determining how resources will be allocated.
5.4.3 Information importance analysis
In this stage, all the sensor/device information received by resource manager will be processed to compute their traffic data values and determine if these traffic data are duplicated or not. The details are as follows:
Deriving sensor/device information
After the vehicle connection establishment stage, the resource manager will receive all the vehicle information, which contains equipped sensors/devices
Figure 4 An example map that shows location blocks.
capabilities, types, and their locations. The resource manager then extracts the sensors/devices information from the vehicle information and stores them separately.
Duplication check and grouping
In this step, the resource manager checks whether there are duplicated traffic data in a certain area. The occurrence of duplicated traffic data implies that there are more than one sensor/device can provide the same type of traffic data within a location block. The resource manager will group them together and choose several of them in each group to allocate resources based on their CQI.
Traffic data value computation
Traffic data value computation is a major step in the information importance analysis stage, which computes the traffic data value of each sensor/device. The value is used to determine whether the traffic data provided by these sensors/devices are important or not. We compute based on vehicle type, sensor/device ability, and location. Each of them corresponds to an importance degree which is derived from the QoS requirements.
The equation below computes the traffic data value based on three
The normalization procedure is used to equalize the influence of each importance channel environment is suitable for transmission.
Required channel resources computation
The resource allocation unit of radio resource in LTE is called a resource block.
The resource manager computes the number of required resource blocks for each sensor/device based on its CQI and sensor/device data producing rate. The cost of traffic data transmission must be measured to evaluate the cost effectiveness of certain CQI, , which also represents the data consuming rate. Based on , we can learn the modulation scheme and code rate to obtain the available data rate according to the LTE specification, which is shown in Table 4 [24]. Note that the larger indicates that we need more resource blocks to support the transmission, therefore, can be used to present the transmission cost.
Sorting for resource allocation priority
Table 4 CQI MCS mapping table in LTE [24].
CQI index Modulation Code rate x 1024 efficiency
0 Out of range
Resource limitation check
This step is used to check whether there are enough radio resource blocks for a certain sensor/device. The checking order is according to the previous sorted result.
Let be the total number of resource blocks and is the remaining number of resource blocks. The number of can be derived from the channel bandwidths [25]. We will sequentially check for each if is enough to allocate or not as follows:
∑
Allocation permitted
After the above procedure, resource manager permits to reserve resources to the vehicular sensors/devices, and decide the QoS parameter settings including QCI, GBR and MBR based on QoS requirements.
5.4.5 Resource reservation
Finally, the resource manager sends the resource allocation results back to the LTE P-GW. Once the dedicated bearer establishment procedure starts, P-GW will inform UEs (vehicles) how resources are allocated based on the previous results.
Chapter 6
Performance Evaluation
6.1 Experimental setup
We implement a system level simulation for our resource management algorithm for multimedia traffic report systems (QoS-MTRS). The environment we built is under the open source network simulator, NS3 [10]. The simulation parameters show in Table 5 below. We simulated 50 eNodeBs, 100 UEs (vehicles) and 200 sensors distributed in a map with 24/54/100 location blocks. The map and the movement trace file of UEs are generated via VanetMobiSim [11].
Table 5 Simulation parameter settings [12][13].
Parameter Value
Mobility model IDM_LC (by VanetMobiSim) RandomWayPoint
Simulated areas 0.3 x 0.2 / 0.4 x 0.5 / 1 x 1 km2
Topology Urban/suburban
Simulation time 5 / 10 / 30 s
Table 6, Table 7 and Table 8 are the importance degrees we adopted in the simulations. Since we regard importance degrees as given values, the numerical values of them are generated randomly.
Table 7 Importance degrees of vehicles.
Vehicle Type 𝒅𝒊𝒗
Police/ Ambulance/ Fire truck 3
Public transportation 2
Household vehicle 1
Table 6 Importance degrees of sensors/devices.
Data Type Data Format Sensor /
6.2 Evaluation results of QoS-MTRS
For performance evaluation, we compare the proposed QoS-MTRS with BQA (bandwidth and QoS-aware scheduler) [5] and two static methods: FCFS (first-come, first-served) and Greedy (greedy, based on traffic data value) scheduler. The performances we concerned are the appearance probability of received traffic data types, overall system throughput, end-to-end delay, duplication ratio (Dup), coverage (Cov) and overall value of received traffic data (S). The above performance metrics have been defined in Chapter 4.
Table 8 Importance degrees of location blocks.
Location
Figure 5 shows the appearance probability distribution of the received sensor/device types per location block by using our proposed QoS-MTRS and other scheduling mechanisms. Let n represents the number of received sensor/device types in a certain block. n = 0 means there is no traffic data reported by any sensor/device.
In our proposed method, the probability of n = 0 is 80% lower than that of any other method on average, and the probability of n > 1 is 65%, 58% and 48% higher than BQA [5], Greedy and FCFS, respectively, representing that the proposed QoS-MTRS has more chances to receive traffic data in various types.
Figure 5 Comparison of appearance probability distributions of the received data types per location block.
Figure 6 presents the comparison of duplication ratio, . The duplicated data may result in resource wasting because we allocate more resources but only received few kinds of data. Hence, low is preferred. Simulation result shows that our proposed method can decrease 50% of the duplicated data on average.
Figure 6 Comparison of duplication ratio.
The overall traffic data value, , (excluded duplicated data) is shown in Figure 7.
It represents the overall traffic data value we received, which means the worth of the received traffic data. Simulation result shows that our proposed QoS-MTRS can get the most valuable traffic data even when duplicated data are discarded. In terms of , the proposed QoS-MTRS is 27%, 18% and 54% higher than BQA [5], Greedy and FCFS, respectively.
Figure 7 Comparison of overall traffic data value.
Figure 8 shows the cumulative distribution function of end-to-end delay. Here, end-to-end means from UE to the traffic report application server in the Internet. We measured the end-to-end delay from the simulated trace file produced by NS3.
Simulation result shows that there are 90% of the end-to-end delay under 0.016 seconds in the proposed QoS-MTRS and BQA [5], while for Greedy and FCFS, there are just 63% of the end-to-end delay under 0.016 seconds.
Figure 8 Comparison of End-to-end delay.
Figure 9 presents the system throughput. The throughput was derived from the simulated packet trace files produced by NS3, and was further processed by the Wireshark software. The stable growth of our proposed method indicates that the resources are allocated properly with channel quality checking. Simulation result indicates that the throughput of the proposed QoS-MTRS is 27%, 27% and 83%
higher than that of BQA [5], Greedy and FCFS, respectively.
Figure 9 Comparison of system throughput.
Figure 10 shows the coverage ( ) comparison among different methods. Our proposed QoS-MTRS always achieves at least 87% coverage even when the number of vehicles is only 20, while BQA [5], Greedy and FCFS just reach 30%, 58% and 64% coverage, respectively.
Figure 10 Comparison of coverage.
Chapter 7
Conclusion and Future Work
7.1 Concluding remarks
In this paper, we have presented a QoS-aware resource management mechanism for multimedia traffic report systems (QoS-MTRS) over LTE. The main idea of the proposed mechanism is that we consider both information importance and channel quality to improve the diversity, completeness and overall traffic data value under certain resource (e.g., radio) limitations. We have conducted a system-level simulation that includes LTE network environment to evaluate our design. Simulation results show that by adopting the proposed QoS-MTRS, the probability of receiving more than one kind of traffic data in a location is 65%, 58% and 48% higher than BQA [5], Greedy and FCFS, respectively; the duplication ratio can be decreased by 50% on average. The overall traffic data value of QoS-MTRS is 27%, 18% and 54% higher than that of BQA [5], Greedy and FCFS, respectively. In addition, QoS-MTRS always achieves at least 87% coverage even when the number of vehicles is as low as 20.
7.2 Future work
In the proposed QoS-MTRS, the importance degrees of vehicle, sensor/device types and locations are given values, and we just focus on the end-to-end resources management in LTE networks. For our future works, we will derive importance degrees from statistical data and integrate the QoS-MTRS with M2M architecture to support resource management between LTE networks and sensor networks, and further port the QoS-MTRS to LTE-A environments. Moreover, the resource
management problem may be modeled as an optimization problem for mathematical analysis.
References
[1] A. E. Essaili, E. Steinbach, D. Munaretto, S. Thakolsri and W. Kellerer,
“QoE-driven resource optimization for user generated video content in next generation mobile networks,” in Proc. 2011 18th IEEE International Conference on Image Processing (ICIP), Sep. 2011, pp. 913-916.
[2] A.G. Gotsis, A. S. Lioumpas and A. Alexiou, “M2M scheduling over LTE:
challenges and new perspectives,” in Vehicular Technology Magazine, IEEE, Vol.
7, pp. 34-39, Sep. 2012.
[3] I. C. Wong, O. Oteri and W. McCoy, “Optimal resource allocation in uplink SC-FDMA systems,” in IEEE Transactions on Wireless Communications, Vol. 8, May. 2009.
[4] R. Kwan, C. Leung, and J. Zhang, “Proportional fair multiuser scheduling in LTE,”
in IEEE Signal Processing Letters, Vol. 16, pp. 461-464, Jun. 2009.
[5] S. N. K. Marwat, Y. Zaki, C. Goerg, T. Weerawardane and A. Timm-Giel,
“Design and performance analysis of bandwidth and QoS aware LTE uplink scheduler in heterogeneous traffic environment,” in Proc. IEEE International
“Design and performance analysis of bandwidth and QoS aware LTE uplink scheduler in heterogeneous traffic environment,” in Proc. IEEE International