Enhanced low latency queuing algorithm with active queue management for multimedia applications in wireless networks
Rukmani Panjanathan* and Ganesan Ramachandran School of Computing Science and Engineering, VIT University Chennai Campus, Vandalur-Kelambakkam Road, Melakottaiyur, Chennai, Tamil Nadu, 600127, India Email: [email protected] Email: [email protected] *Corresponding author
Abstract: Low cost broadband services used widely in the recent years increased the demand for multimedia applications. Many of these applications require different quality of service (QoS) in terms of throughput and delay. In the current resource constrained wireless networks, resource management plays a vital role. In order to handle the resources effectively and to increase the QoS, proper packet scheduling algorithms need to be developed. Low latency queuing (LLQ) is a packet scheduling algorithm which combines strict priority queuing (SPQ) to class-based weighted fair queuing (CBWFQ). In this paper, an enhanced low latency queuing (ELLQ) algorithm is proposed in which an additional SPQ is introduced along with the existing SPQ. Further, the QoS is improved by integrating congestion avoidance algorithm with ELLQ. Simulation results using the OPNET modeller show that the proposed algorithm outperforms the existing algorithm in terms of throughput and delay for the multimedia applications.
Keywords: low latency queuing; LLQ; random early detection; RED; multimedia applications; quality of service; QoS; scheduling algorithms; wireless networks.
Reference to this paper should be made as follows: Panjanathan, R. and Ramachandran, G. (2017) ‘Enhanced low latency queuing algorithm with active queue management for multimedia applications in wireless networks’, Int. J. High Performance Computing and Networking, Vol. 10, Nos. 1/2, pp.23–33.
Biographical notes: Rukmani Panjanathan received her MTech in Information Technology from the Anna University, Tamil Nadu in 2007. She is currently working as an Assistant Professor and pursuing her PhD in the School of Computing Science and Engineering, VIT University, Chennai Campus. Her research interests include QoS for real-time applications and scheduling algorithms for wireless networks.
Ganesan Ramachandran received his PhD in Information System Security from Bharathiar University, Tamil Nadu in 2010. He is currently working as an Associate Professor in the School of Computing Science and Engineering, VIT University, Chennai Campus for the past three years. He is having more than 15 years of teaching experience. He is an active member of IA Eng., CSI, IACSIT and CRSI. He is the technical and review committee member of various international conferences and international journals. He has published research articles in national and international journals and conferences. His area of specialisation is network security and information security.
1 Introduction Fast development in the high speed low cost wireless communications had increased the demand for multimedia applications in various domains. Video applications are gaining widespread popularity among various territories. Remote surgery (telesurgery) in healthcare domain is one such territory which is becoming more and more popular with the advancements in multimedia. Another territory is the e-learning sector where numerous interactive web-based knowledge sharing systems are developed by many
institutions incorporating the media streaming applications in major level (Jameel et al., 2014).
Best quality of service (QoS) on high quality video transmission along with the end to end performance is the need of this hour for these applications (Hua-Ching et al., 2013). Assessments of the performance of these multimedia applications are provided in terms of the quality of experience (QoE). Video QoE effects are caused by the QoS problems such as bandwidth, jitter, delay, loss and
24 R. Panjanathan and G. Ramachandran
throughput. Increasing the QoS of these applications guarantees the user satisfaction.
However, provision of high quality audio and video applications over the wireless networks is a great challenge due to the distinctive nature of inadequate resources and time-varying channel state (Jian and Ye-Qiong, 2009; Peng et al., 2011). Lack of guaranteed QoS may make the users unsatisfied with the services provided by their service providers (Jameel et al., 2014). Guaranteed QoS can be offered by managing the buffer space in routers efficiently. This can be achieved by the packet scheduling algorithms at layer 3. The main job of the packet scheduling algorithm is to determine the packet transmission order among different service classes of traffic which directly affects throughput, delay and loss rate (Nazy and Ljiljana, 2002).
The traditional packet scheduling algorithm is the first-in-first-out (FIFO). FIFO places all the packets into a single queue and processes them in the same order as they are arrived. This is easy to implement but the drawback is it cannot differentiate the types of traffics. For instance, if a bursty traffic comes in, FIFO will use the whole buffer which may cause delay in the real-time sensitive traffic. Also, other flows may not be serviced until the buffer is empty (Tetsuji, 2010). To overcome these limitations and to provide fair sharing of the resources, many other types of scheduling methods such as priority queuing (PQ), weighted round robin (WRR), weighted fair queuing (WFQ), custom queuing (CQ) and class-based weighted fair queuing (CB-WFQ) are designed.
Real-time applications are treated preferentially by the PQ algorithms. But, when the amount of high priority traffic is excessive, PQ suffers starvation problem and complete resource malnourishment for the low priority traffic (Tetsuji, 2010). In case of WRR, WFQ, CQ and CBWFQ, there is no priority servicing for the real-time applications. In order to overcome these problems, Cisco Systems has introduced a low latency queuing (LLQ) algorithm which combines a single strict priority queue (SPQ) with CBWFQ (Low Latency Queuing, http://www.cisco.com/c/en/us/td /docs/ios/12_0s/feature/guide/fsllq26.html). LLQ places delay sensitive voice application in the SPQ and treat it preferentially over other traffic by scheduling the voice application first before the packets in the other queues (Brunonas et al., 2006). The key difference between LLQ and PQ is that the LLQ SPQ will not starve the low priority queues. The SPQ is controlled by the policer at the time of congestion either by the bandwidth or a percentage of the bandwidth (Chuck, 2001). But in the traditional PQ, there is no such controller to avoid the starvation problem. So, the SPQ in the LLQ could not consume more bandwidth than the assigned bandwidth. Other network corporates like FatPipe Networks (http:// www.fatpipeinc.com/technology/bandwidth-management. php), Juniper Networks (https://www.juniper.net/techpubs/ en_US/junos-space11.4/junos-space-QoS-Design-subindex. html), CheckPoint (https://sc1.checkpoint.com/documents/ R76/CP_R76_QoS_AdminGuide/index.html), and Palo Alto Networks (https://paloaltonetworks.com/documentation/61/
pan-os/pan-os/quality-of-service.html) also sensitise on assigning the high priority to real-time applications.
Traffic congestion has become a severe problem in wireless networks due to heavy traffic. This can melodramatically spoil the performance of the network (Geyong and Xiaolong, 2013; Shihang et al., 2008). The router holds the packets in queues for transmission that are watched by queue management techniques for manifestation of forthcoming congestion. Tail-drop is a standard queue management policy which can be used in internet routers to implement packet dropping and to control the congestion (Shu-Gang, 2008). With the help of tail-drop technique, the queue is permitted to receive packets until it reaches maximum capacity. After this, incoming packets will be rejected until some buffer space becomes available again. For interactive multimedia applications, tail-drop is unsuitable with the constraint of low delay and more throughputs.
In order to overcome this problem, active queue management (AQM) has been proposed which acts as an effective policy of congestion control to attain maximum system utilisation and minimum packet delay. In order to notify the presence of congestion at the initial stage, AQM discards the incoming packets before the existence of the buffer overflow. The most commonly deployed AQM technique is random early detection (RED) (Ivan et al., 2010). Based on the predefined threshold and dropping probability, RED algorithm watches the queue length to drop the incoming packets. But the foremost problem of the original RED algorithm is the absence of differentiated services which is mandatory for multimedia applications. Geyong and Xiaolong (2013) are also addressing the same that the advanced multimedia applications need stringent QoS whereas the traditional non-real-time applications may not require strict QoS. Pure RED does not accommodate this QoS differentiation. Weighted random early detection (WRED) can provide early detection with QoS consideration. In the WRED, different drop probabilities for different priorities can be given based on the IP precedence or DSCP values.
In this paper, an enhanced LLQ algorithm is proposed which extends the structure of the LLQ algorithm. In addition to this, DSCP-based WRED is configured to prevent the congestion and to increase the QoS for multimedia applications.
2 Related work Tedi and Jason (2002) state that demand for multimedia applications got increased owing to rapid growth of internet. Users with multiple applications such as e-mail, file transfer, VoIP and video applications require different levels of QoS and different access rates based on the requirement of the application. With the help of an optimal congestion control, the traffic flow can be regulated to provide the required QoS to the users. There are three ways to provide QoS and to control congestion as listed below:
Enhanced low latency queuing algorithm with active queue management for multimedia applications 25
• Call acceptance control which is an end-to-end mechanism that functions at the two ends of a connection.
• Shaping and policing which is an edge mechanism that works at user-network interface.
• Buffering, queue management and scheduling which are core mechanisms that are functioning at network switching nodes like routers and switches.
Huge amount of bandwidth is required for teleconferencing and video on demand services in internet. Bandwidth provision in large quantities is not an easy task in the internet services due to persistent changing nature of the internet. With the help of an appropriate buffer handling mechanisms, the bandwidth can be managed efficiently (Farzad et al., 2008). Hence, the multimedia applications can be delivered with the required QoS.
Jesus et al. (2006) have developed a model to give more priority to voice and video traffics which are most sensitive. The model monitors all incoming traffics and categorises it based on the level of their sensitivity. Then, it assigns highest priority to voice and video traffic and lower priority to other traffic which are delay tolerant. The prioritised traffic is allocated with the total bandwidth. So, that it can be delivered to the destination directly without considering congestion avoidance technique. But, the low priority traffics follow and share the bandwidth based on some specific set of policies configured for each class of traffic. The experimental results proved that the quality and performance of the voice and video traffic is increased by 70% to 130% when it is being prioritised (Jameel et al., 2014).
Shaimaa et al. (2011) have demonstrated that the combination of class-based weighted fair queuing and low latency queue (CBWFQ-LLQ) improved the performance of multimedia applications. This is verified with the help of experimental results done in OPNET IT Guru as below:
a Without using any QoS technique, the first experiment was conducted on a first come first serve (FIFO). Non-real-time applications such as, Hyper Text Transfer Protocol (HTTP), File Transfer Protocol (FTP), etc. got more privileges and occupied more buffer spaces and introduced negative effect on the multimedia applications, i.e., voice and video as they are waited long and started to drop packets.
b The second experiment was conducted using CBWFQ-LLQ at routers for video traffics. The delay sensitive applications provided with strict priority. The results were shown that the performance of the video traffic has improved but it was affecting other traffics including voice.
c The third experiment prioritises voice and video traffics. It improved the overall performance of the network. To attain even better quality for multimedia applications, the QoS has to be extended to layer 2.
Applying congestion management techniques will improve its overall performance (Jameel et al., 2014).
Geyong and Xiaolong (2013) communicate that congestion control has become one of the most serious issue in recent communication systems. A set of queues which hold packets for transmission are maintained at the internet routers. The queues are watched for occurrence of forthcoming congestion. Tail-drop is a traditional queue management policy which is unsuitable for delay sensitive multimedia applications. The RED algorithm is the most commonly used AQM mechanism (Su et al., 2012). It drops the incoming packets based on the predefined threshold values. The packet dropping probabilities are varied by monitoring the queue length. If there is free space in the buffer, all inward packets are permitted to enter the queue. The packet dropping probability increases when the queue length increases. The packets are dropped when the queue length exceeds the predefined threshold values. When compared to tail-drop, RED is better and it has the potential to overcome some of the problems occurred in the tail-drop policy (Thomas et al., 2000).
The serious drawback in the original RED algorithm is the absence of a differentiated QoS which is mandatory for multimedia applications (Geyong and Xiaolong, 2013). Integrating priority scheduling with RED has been recognised as a vital idea for differentiated services and congestion control. In this combined scheme, the buffer space is separated into different partitions by certain thresholds for various traffic classes. If queue size goes beyond the predefined threshold value of a particular traffic class, the packets from higher priority classes can only be admitted. With this method, it is possible to provide QoS differentiation since higher threshold queue packets are prioritised.
The LLQ is essentially a CBWFQ combined with a SPQ. Traffic assigned to the SPQ is completely serviced before all other CBWFQ queues are serviced. The existing LLQ algorithm gives strict priority mostly to voice applications only. Advancements in the field of telesurgery, video conferencing and e-learning applications had increased the demand for high quality video services in the recent years. Table 1 (Chris and Steve, 2006) shows the QoS requirements for interactive video and voice applications. Interactive video applications, which deliver both video and voice, are very similar to the voice applications in terms of the QoS requirements as depicted in Table 1. The demand for the streaming videos is getting increased these days due to its wide usage in various fields such as e-learning and video on demand. But the streaming video applications are little bit delay and buffer tolerant compared to the interactive video applications. In addition to the voice traffic, strict priority can be applied for the video traffic to satisfy the QoS requirements in interactive video
26 R. Panjanathan and G. Ramachandran
applications and to provide the service guarantee in streaming video applications.
Figure 1 shows the general classification of real-time and non-real-time applications (Quality of Service Design Overview, http://www.cisco.com/c/en/us/td/ docs/solutions/Enterprise/WAN_and_MAN/QoS_SRND/QoS-SRND-Book/QoSIntro.html). In the LLQ, it is possible to include various types of real-time traffic into the single SPQ. But, the expected QoS level cannot be guaranteed, if sensitive audio and video packets are processed in the single SPQ. The reason is that the behaviour of the voice traffic is controllable whereas the video traffic is uncontrollable. Also, to avoid jitter, voice traffic requires a non-variable delay which is the most important for voice applications. But the video traffic could introduce variation in delay, thereby spoiling the steadiness of the delay required for successful voice traffic transmission. And also if a bursty video packet comes, the voice traffic also may not be delivered successfully (Brunonas et al., 2006).
Table 1 QoS requirements
Traffic Loss One way latency
One way jitter
Voice ≤ 1% ≤ 150 ms < 30 ms Interactive video ≤ 1% ≤ 150 ms < 30 ms Streaming video ≤ 5% ≤ 5 sec -
In order to overcome these problems, an additional SPQ is introduced along with the existing SPQ in the proposed system as illustrated in Figure 2. The existing SPQ which is primarily dedicated for delay sensitive voice traffic in the existing algorithm is re-named as primary strict priority queue (PSPQ) and a new queue is added which is named as secondary strict priority queue (SSPQ). This SSPQ will be exclusively used for video traffic. The SSPQ is configured with a DSCP-based WRED (D-WRED) congestion avoidance algorithm. This D-WRED is an extended WRED which supports differentiated services (DiffServ) between the interactive and streaming video traffics. The SSPQ is monitored by the D-WRED which is configured to drop streaming video packets when the queue size reaches 50% of the total length. The reason behind dropping the streaming video traffic is to create more space for interactive video traffic which is more delay sensitive than streaming video applications and to increase the QoS. Entertainment videos will be classified as non-organisational applications. These video applications may be marked as scavenger class (DSCP CS1) and they will be assigned with minimal percentage of bandwidth. Also, these video applications may not be processed at the time of congestion.
Figure 1 11-class QoS baseline queuing model (see online version for colours)
All other classes of traffic are processed using CBWFQ algorithm with default dropping method. When the TX Ring (hardware queue) has free space, voice packets will be scheduled from PSPQ first and then the video packets will be scheduled from the SSPQ. When both queues are empty the packets from other queues will get an opportunity to be processed based on the CBWFQ algorithm. At the time of congestion, the PSPQ and SSPQ will be monitored by the bandwidth policer.
The algorithm for the CA-ELLQ is given as follows:
a packet classifier will classify incoming packets based on type of service (TOS) field
b voice and video packets will be placed into the PSPQ and SSPQ, respectively
Enhanced low latency queuing algorithm with active queue management for multimedia applications 27
c other type of packets will be placed into the respective queues based on CBWFQ algorithm
d when the TX Ring (hardware queue) has free space, packets from PSPQ will be processed first
e when PSPQ is empty, packets from SSPQ will be processed
f when PSPQ and SSPQ are empty, packets from other queues will be processed based on CBWFQ
g when congestion occurs, the Voice and Video policer will monitor the PSPQ and SSPQ
h the policer will drop the incoming voice and video packets if the allotted bandwidth for the SPQ goes beyond its maximum limit.
4 Simulation settings Simulation results are provided to show the performance of the proposed algorithm using OPNET Modeler (Version 14.5) (Adarshpal and Vasil, 2013). Our simulation model has been created using 30 mobile nodes, six Ethernet routers, two Ethernet switches and 2 CISCO7200 gateway routers in a 10 × 10 km area as shown in Figure 3. The mobile nodes will be moving based on the predefined trajectory.
Figure 3 Network model in OPNET simulator (see online version for colours)
Input traffic models for voice and video applications are created using application and profile configuration as shown in Figure 4. In the application definition, TOS used to represent voice and video traffic is interactive voice,
streaming and interactive multimedia. The corresponding DSCP values for the created input traffic models are also assigned in the application definition. Then, the input traffic models will be assigned to the mobile nodes with the help of application profiles. The D-WRED profile is created as shown in Figure 5.
Figure 4 Profile configuration (see online version for colours)
Figure 5 D-WRED configuration (see online version for colours)
28 R. Panjanathan and G. Ramachandran
Figure 6 Queuing rules for 11-class queuing models (see online version for colours)
Two scenarios have been designed to evaluate the performance of the proposed algorithm along with the existing algorithm. The LLQ and the proposed CA-ELLQ are analysed in scenarios 1 and 2, respectively. The gateway router’s (CISCO7200) buffer space is configured based on the guidelines given in Figure 6.
5 Simulation results analysis The performance of the proposed algorithm is analysed by comparing with the LLQ algorithm in terms of throughput and delay for the voice and video applications. Each factor is discussed elaborately with graphical results in the following sections. To better understand the performance of the proposed system, statistical results obtained through simulations also shown in Table 2.
Table 2 Simulation results statistics
Algorithm LLQ CA-ELLQ
Average
Voice traffic sent (bytes/sec) 3813 3,976.3 Voice traffic received (bytes/sec) 3810 3,972.8 Voice packet end-to-end delay (sec) 0.080929 0.082044 Voice jitter (sec) 0.0001236 0.0000690 Video traffic sent (bytes/sec) 449,280 449,280 Video traffic received (bytes/sec) 341,338 386,035 Video packet end-to-end delay (sec) 0.040370 0.044867 WLAN data dropped (buffer overflow) (bits/sec)
5.1 Voice traffic sent It is observed from the graphical results shown in Figure 7, that average voice traffic sent in the LLQ is 3,813 (bytes/sec) and in the CA-ELLQ it is 3,976.3 (bytes/sec). Therefore, the overall traffic sent is increased by 4.1% in the proposed algorithm.
Figure 7 Voice traffic sent (bytes/sec) (see online version for colours)
The dedicated PSPQ for the voice applications in the CA-ELLQ improved the overall traffic sent compared to the LLQ algorithm, where both voice and video applications share the single SPQ. Increase in the voice traffic sent is
Enhanced low latency queuing algorithm with active queue management for multimedia applications 29
very important in determining the performance of our proposed work. The dedicated PSPQ gives a notable increase in the voice traffic sent which helps in providing a good quality voice and henceforth increases the total amount of voice traffic to be sent.
5.2 Voice traffic received By looking at the graphical results shown in Figure 8, it can be evident that the average voice traffic received in the LLQ is 3,810 (bytes/sec) and in the CA-ELLQ it is 3,972.8 (bytes/sec). It shows 4% increase in the proposed system. The increase in the value of the voice traffic received shows that the voice packets are being received continuously at the receiver side with good quality.
Figure 8 Voice traffic received (bytes/sec) (see online version for colours)
5.3 Voice traffic end-to-end delay The graphical results in Figure 9 show the end-to-end delay obtained for the voice traffic. An average delay of 0.082044 (sec) is obtained in the CA-ELLQ compared to the 0.080929 (sec) average delay of LLQ. This indicates a slight increment in the delay in the proposed algorithm.
The slight variation in the delay had occurred due to sharing of the total bandwidth assigned to the single priority queue by the two priority queues in the proposed algorithm. But this delay is below the acceptable level based on the QoS requirements as shown in Table 1 for the voice applications. Adaptive scheduling between the SPQs can be considered to reduce the delay.
5.4 Voice traffic jitter The simulation results shown in Figure 10 give the average jitter for voice traffic. LLQ value is 0.0001236 (sec) whereas the CA-ELLQ gives 0.0000690 (sec). The numerical results show that the proposed system gives much less jitter than the LLQ. The decrease in the value of jitter increases the voice quality and the amount of traffic to be received.
Figure 9 Voice traffic end-to-end delay (sec) (see online version for colours)
Figure 10 Voice jitter (sec) (see online version for colours)
30 R. Panjanathan and G. Ramachandran
5.5 Video traffic sent It is observed from the graphical results shown in Figure 11, that average video traffic sent in the LLQ is 449,280 (bytes/sec) and in the CA-ELLQ is 449,280 (bytes/sec).
Figure 11 Video traffic sent (bytes/sec) (see online version for colours)
The router’s buffer space is configured with 75% of the available interface bandwidth. In this, 33% of the bandwidth is allocated for the SPQ. In the proposed algorithm, there is no additional bandwidth for the newly added priority queue. The existing 33% is shared between the PSPQ and SSPQ. The ratio is based on the designer’s choice. In our network scenario, we have given equal share between the priority queues. Therefore, the proposed algorithm processes the incoming video traffic as similar to the existing algorithm.
5.6 Video traffic received By looking at the graphical results shown in Figure 12, the average video traffic received in the LLQ is 341,338 (bytes/sec) and in the CA-ELLQ is 386,035 (bytes/sec). The overall traffic received is increased by 11% in the proposed algorithm.
The dedicated congestion aware SSPQ for the video traffic in the CA-ELLQ improved the overall traffic received than the LLQ algorithm. Increase in the video traffic received is very important in determining the performance of our proposed work for video applications. The newly added SSPQ gives a notable increase in the video traffic received which helps in providing a good quality video and also increases the total amount of video traffic to be received.
Figure 12 Video traffic received (bytes/sec) (see online version for colours)
5.7 Video traffic end-to-end delay The graphical results in Figure 13 show the end-to-end delay obtained for the video traffic. The LLQ gives the average delay as 0.040370 (sec) and the CA-ELLQ gives 0.044867 (sec) as average delay. The proposed algorithm gives slight increment in the delay.
Figure 13 Video traffic end-to-end delay (sec) (see online version for colours)
Enhanced low latency queuing algorithm with active queue management for multimedia applications 31
The slight variation in the delay might have occurred because the total bandwidth assigned to the single priority queue has been shared by two priority queues in the proposed algorithm. But this delay is below the acceptable level based on QoS requirements as shown in Table 1 for the video applications.
5.8 WLAN data dropped The graphical results shown in Figure 14 give the WLAN data dropped due to buffer overflow. The data dropped in LLQ is 101,403 (bits/sec) and in the CA-ELLQ it is 56,012 (bits/sec). There is a drastic decrease in the overall traffic dropped owing to buffer overflow in the proposed algorithm.
Figure 14 Wireless LAN data dropped (buffer overflow) (bits/sec) (see online version for colours)
The reason behind the drastic decrease in data drop in the network is the introduction of an additional SSPQ. Both voice and video traffic are serviced separately in two different queues, thereby reducing the overall data drop rate.
5.9 WLAN delay By looking at the graphical results shown in Figure 15, the WLAN delay in LLQ is 0.029548 (sec) and in the CA-ELLQ it is 0.028705 (sec). The overall WLAN delay is reduced by 3% in the proposed algorithm. It is observed from the graphical results that the proposed system gives good improvement in the overall network delay.
Additional SSPQ introduced in the proposed algorithm results in the drastic decrease in the overall network delay. Both voice and video traffic are serviced separately in two different queues thereby it reduced the overall delay.
Figure 15 Wireless LAN delay (sec) (see online version for colours)
5.10 WLAN throughput The two graphical results shown in Figure 16 give the WLAN throughput obtained in the LLQ and the CA-ELLQ algorithm. The throughput obtained in LLQ is 3,592,922 (bits/sec) and in the CA-ELLQ it is 3,639,007 (bits/sec). There is a drastic increase in the overall throughput in the proposed algorithm.
Figure 16 Wireless LAN throughput (bits/sec) (see online version for colours)
32 R. Panjanathan and G. Ramachandran
Throughput is the most important measurement to determine the performance of the proposed system. It is very difficult to get equal throughput in any network. But the objective of our proposed system is to increase the throughput for multimedia applications. The voice and video traffic are serviced separately in two different queues thereby the proposed system increased the overall network throughput by 1.2%.
6 Conclusions In this paper, we proposed a congestion aware enhanced low latency queuing (CA-ELLQ) algorithm that effectively supports real-time and non-real-time applications over wireless networks. The CA-ELLQ algorithm uses two SPQs namely PSPQ for voice and SSPQ for video traffic which is configured with DSCP-based WRED. Through the simulation results, it is shown that the proposed algorithm guarantees maximum throughput and satisfactory end-to-end delay towards delay sensitive voice and video traffic when compared to the existing LLQ algorithm. Also, it is observed that the end-to-end delay is slightly higher in the proposed algorithm.
As a future work, it has been planned to develop some appropriate strategies to overcome these issues for the voice and video traffic as an extension to the proposed algorithm. Also, it could be possible to schedule the video packets from the SSPQ before scheduling the packets from the PSPQ by considering the nature of the application. And, an appropriate method also can be developed to decide the selection of the SPQs for scheduling the packets dynamically.
Acknowledgements We thank Riverbed for the OPNET Modeler 14.5, freeware released for academic research purpose.
References Adarshpal, S.S. and Vasil, Y.H. (2013) The Practical OPNET
User Guide for Computer Network Simulation, 1st ed., CRC Press, New York.
Brunonas, D., Tomas, A. and Aurelijus, B. (2006) ‘Analysis of QoS assurance using weighted fair queueing (WFQ) scheduling discipline with low latency queue (LLQ)’, 28th International Conference on Information Technology Interfaces, pp.507–512.
Checkpoint [online] https://sc1.checkpoint.com/documents/R76/ CP_R76_QoS_AdminGuide/index.html (accessed 3 August 2015).
Chris, L. and Steve, P. (2006) ‘Implementing quality of service over Cisco MPLS VPNs’, Selecting MPLS VPN Services, Cisco Press.
Chuck, S. (2001) Supporting Differentiated Service Classes: Queue Scheduling Disciplines, White Paper, Juniper Networks.
Farzad, A., Sahar, K. and Bahram, S.B. (2008) ‘A new scheduling algorithm based on traffic classification using imprecise computation’, International Journal of Computer, Control, Quantum and Information Engineering, Vol. 2, No. 9, pp.78–82.
Fatpipe Networks [online] http://www.fatpipeinc.com/technology/ bandwidth-management.php (accessed 3 August 2015).
Geyong, M. and Xiaolong, J. (2013) ‘Analytical modelling and optimization of congestion control for prioritized multi-class self-similar traffic’, IEEE Transactions on Communications, Vol. 61, No. 1, pp.257–265.
Hua-Ching, C., Hsuan-Ming, F. and Dong-hui, G. (2013) ‘Novel MPEG-4 video cross-layer algorithm in wireless networks’, International Journal of Computational Science and Engineering, Vol. 8, No. 2, pp.171–179.
Jameel, S.Y., Suhaimi, B.A.L., Aisha, H.A.H. and Korshed, A. (2014) ‘An improved QoS in the architecture, model and huge traffic of multi-media applications under high speed wireless campus network’, ARPN Journal of Engineering and Applied Sciences, Vol. 9, No. 12, pp.2780–2793.
Jesus, A., Perez, V.H.Z. and Christian, C. (2006) ‘A network and data link layer QoS model to improve traffic performance’, International Conference on Emerging Directions in Embedded and Ubiquitous Computing, pp.224–233.
Jian, L. and Ye-Qiong, S. (2009) ‘DLB: a novel real-time QoS control mechanism for multimedia transmission’, International Journal of High Performance Computing and Networking, Vol. 6, No. 1, pp.4–14.
Juniper Networks [online] https://www.juniper.net/techpubs/ en_US/junos-space11.4/junos-space-QoS-Design-subindex.html (accessed 3 August 2015).
Low Latency Queuing [online] http://www.cisco.com/c/en/us/td/ docs/ios/12_0s/feature/guide/fsllq26.html (accessed 18 June 2015).
Nazy, A. and Ljiljana, T. (2001) ‘Implementation of VirtualClock scheduling algorithm in OPNET’, Proceedings of OPNETWORK, pp.1–7.
Palo Alto Networks [online] https://paloaltonetworks.com/ documentation/61/pan-os/pan-os/quality-of-service.html (accessed 3 August 2015).
Peng, L., Yilin, C., Nina, F. and Fuzheng, Y. (2011) ‘A cross-layer algorithm of packet scheduling and resource allocation for multi-user wireless video transmission’, IEEE Transactions on Consumer Electronics, Vol. 57, No. 3, pp.1128–1134.
Quality of Service Design Overview [online] http://www.cisco.com/c/en/us/td/docs/solutions/Enterprise/ WAN_and_MAN/QoS_SRND/QoS-SRND-Book/QoSIntro.html (accessed 18 June 2015).
Shaimaa, B., Fatma, B. and Gamal, D. (2011) ‘QoS adaptation in real time systems based on CBWFQ’, 28th National Radio Science Conference, pp.1–8.
Shihang, Y., Geyong, M. and Irfan, A. (2008) ‘Effective admission and congestion control for interconnection networks of cluster computing systems’, International Journal of High Performance Computing and Networking, Vol. 5, Nos. 5/6, pp.374–380.
Shu-Gang, L. (2008) ‘Simulation and evaluation of random early detection in congestion control’, Proceedings of the Seventh International Conference on Machine Learning and Cybernetics, pp.2363–2368.
Enhanced low latency queuing algorithm with active queue management for multimedia applications 33
Su, G., Pham, T.O. and He, S. (2012) ‘QoS guarantee for IPTV using low latency queuing with various dropping schemes’, International Conference on Systems and Informatics, pp.1551–1555.
Tedi, S. and Jason, S. (2002) Smart Queuing: An Adaptive Approach High-Performance Networks, Final Project, Spring 2002, School of Engineering Science, Simon Fraser University, Burnaby [online] http://www2.ensc.sfu.ca/ ~ljilja/ENSC835/Spring02/Projects/susanto_sze/Smart_report.pdf (accessed 18 June 2015).
Tetsuji, H. (2010) ‘Analysis of multiclass feedback queues and its application to a packet scheduling problem’, Journal of Industrial and Management Optimization, Vol. 6, No. 3, pp.541–568.
Thomas, S., Martin, M. and Hannuksela, W. (2000) ‘Analytic evaluation of RED performance’, Nineteenth Annual Joint Conference of the IEEE Computer and Communications Societies, pp.1415–1424.
