Comparison study of FIFO and MDRR queuing ... - IEEE Xplore

0 downloads 0 Views 722KB Size Report
networks: First-in-First-out (FIFO) and Modified Deficit. Round-Robin (MDRR). Simulation results demonstrate that. MDRR has better performance over FIFO in ...
2018 9th International Conference on Information and Communication Systems (ICICS)

Comparison Study of FIFO and MDRR Queuing Mechanisms On 5G Cellular Network Mohammad M. Shurman1, Reem M. Al-Rashdan2, Majeda K. Al-Bataineh2 1Jordan

University of Science and Technology/ Network Engineering and Security Department, Irbid, Jordan University of Science and Technology/ Computer Engineering Department, Irbid, Jordan [email protected], {rmalrashdan15, mkalbataineh15}@cit.just.edu.jo

2Jordan

Abstract – Queueing model behavior affects 5G mobile cognitive radio networks performance, in term of devices lifetime and energy conservation. Maximization of the lifetime and reduction of energy consumption by devices are directly affected by the overall delay, including the queuing delay. In this work, we study two types of queues used in 5G networks: First-in-First-out (FIFO) and Modified Deficit Round-Robin (MDRR). Simulation results demonstrate that MDRR has better performance over FIFO in most metrics, such as end-to-end delay, average ethernet delay (sec), receiver utilization and receiver throughput. Keywords: 5G, OPNET Simulator, FIFO Queue, MDRR

Queue. I.

INTRODUCTION

The demand for high throughput and low delay for multimedia applications in mobile nodes creates the need for 4G wireless communication systems. 4G systems have many improvements over the previous mobile generations. Later on, with the increasing demand for higher throughput, lower delay, the existence of powerful mobile nodes and higher number of users, the demand for newer system has emerged. Thus, researchers around the world is collaborating to create a new generation of wireless communication systems (5G). 5G generation is proposed to accommodate challenges that 4G could not unravel, such as spectrum crisis, highenergy consumption, higher data rate and high mobility. For such technology, with very high expectations and services never seen before, several requirements 5G must achieve, such as system capacity, coverage, availability, higher data rate, reduction in energy usage, and latency reduction [1]. The vision behind 5G technology is the provision of unlimited data and services for everyone using the evolved radio access technologies by enhancing and completing the existing wireless communication technologies with emerging radio access technologies for specific network deployment and usage [2]. In 5G mobile network, there is an ability to exploit under-utilized radio spectrum portions, such as TV white space (TVWS). TVWS includes a large portion of available regional level radio spectrum to support the design and deployment of low power and low cost mobile system on propagation conditions. To reduce the investment cost and provide low priced services, 5G mobile network is deployed over TVWS to enable network operators to cover large geographical areas with lower number of base stations [4].

978-1-5386-4366-2/18/$31.00 ©2018 IEEE

One of the most important issues facing the design of 5G mobile networks is energy efficiency issue. The formation of such emerging technology implies huge energy consumption in the operator’s systems and users’ mobile devices. To meet the optimal energy consumption in the endto-end path, an efficient methodology should combine the temporal bandwidth aware behavior of the node with access policy using both mobile peer exchange mode and radio access mode [4]. The motivitions behind this paper is to shed a glimpse of light on two queuing systems suggested for 5G to determine the effects of such models on overall delay and energy consumption to determine the better one to be used in 5G networks The rest of this paper is organized as follows, section II describes queuing methods related works. Section III presents the methodology for our work, while section IV explains experiments and simulation results. Section V concludes our work. II.

RELATED WORK

Toward the increasing interest in the deployment of 5G technology in the network, several issues should be addressed, considering different perspectives such as the mass increase of data traffic congestion, and the number of connected mobile devices. Increasing the network lifetime is a critical design issue for 5G mobile network technologies. Network nodes operate on limited energy resources; thus, the main objective is to maximize their lifetime. Hence, energy consumed by mobile devices and resources management framework poses a great challenge for wireless access technology planning developers. As proposed in [4], 5G mobile cognitive radio network architecture shows efficient TVWS exploitation and maximum energy conservation. This architecture includes first-in-first-out (FIFO) queuing delay in the measurement of the overall delay, with system consists of three submodules, spectrum broker that coordinate TVWS management, 5G mobile base stations, and radio resources management module in charge of matching requests from 5G base station with the available TVWS, according to the quality of service requirements. The optimum energy-efficient path during the resource exchange requests is the responsibility of the energy controller and it is measured through the streaming path lifetime maximization model (SPLM). SPLM scheme used to keep network nodes alive as long as possible by evaluating each path using tables for the remaining energy on each

61

2018 9th International Conference on Information and Communication Systems (ICICS)

node to choose the optimal path for resource sharing process [4]. When a request from the 5G base station is initiated, the spectrum broker informs the 5G base station about the available parts of TVWS and the price of these available parts, including spectrum portfolio and price portfolio. Next, the 5G base station sends a request about the available radio spectrum portions to the spectrum broker to collect them in the radio resources management module maximizing node’s lifetime and reducing delay [4]. Authors of [5] estimate queuing delay distribution on the internet link carrying TCP traffic, caused by transmitting a huge number of files in long tail distribution connections . Three queuing methods were tested in [6]: FIFO, Weighted Fair Queuing (WFQ) and Priority Queuing (PQ). Results based on network routing, routing delay, network throughput and CPU/memory usage prove that PQ does not need high specification on memory or CPU. It was noted that this mechanism is unfair, it serves one application at a time increasing the delay and FIFO has a smaller delay than the PQ. Using device-to-device (D2D) communications by redefining the existing resources with efficient scheduling methodology, higher gains can be obtained as [7]. In this work, they proved that using Round Robin algorithm (RR) enhances the channel efficiency and accomplishes higher transmission rates. In [8], Weighted Round Robin (WRR) to minimize total end-to-end delay in the system on chip (SoC) compared with RR was used. Their system is based on network calculus method [9]. They proved that WRR has a better delay than RR. However, considering that 5G mobile networks adaptation is associated with energy consumption is proposed in [10] and [11]. Energy efficiency as a vital issue towards low energy consumption levels and low network operation cost, even with a massive increase of data traffic was proposed in [12]. The main characteristics of a 5G mobile network technology to support up to 25Mb/scene is presented in [13]. For wireless networks throughput and delay enhancement using RR method with time-since-last-service (TSLS) counter optimal performance is achieved an excellent delay is discussed in [14]. The proposed works in [15] and [16] provide detailed analysis on energy consumption reduction of mobile devices and the increase of inter-connected devices in the mobile access network. III. METHODOLOGY In this section, our work is based on a system of 5G mobile cognitive radio network architecture for maximum energy conservation. For high energy conservation in any network system, the conservation should be applied to each node in that system. On the node level, the most important value to maximize in order to accomplish that goal is the node lifetime, where the network lifetime is the time from network start to the time to the depletion of the first node energy. Streaming path lifetime maximization (SPLM) scheme is a concept implemented by [4] to estimate the energy efficiency of each energy model subsystem by the allocation of the optimal energy aware path, either through radio resource management module or peer-to-peer resource exchange process. After the system reserved the available

TVWS spectrum with the optimal path and the specific constraint as the node request.

Fig. 1: Streaming Path.

The measurement of the total power, lifetime, and energy consumed in each node in the path includes the queuing delay, based on the incoming traffic and the average service time using FIFO approach as in [3]. For example, if we assume that the path from node A to node D is the optimal path chosen by the system as in Fig. 1. Thus, the totals queuing delay is the summation of delay in node B (δNB), node C (δNC) and from node D (δND). Our study based on the fact that the node lifetime is inversely proportional to the value of queuing delay, so to minimize the queuing delay we used MDRR approach rather than FIFO. FIFO queuing has a basic principle that the first packet arrives at a router is the first packet to be transmitted. We used MDRR as a scheduling algorithm for the network, hoping for more fairness at scheduling the nodes in the queue. MDRR scheduler handles N flows, each queue has several bytes that can transmit at each turn and the remaining if any, is reported to the next turn and that number called quantum. First, MDRR scans all nonempty queues in sequence, the algorithm keeps an auxiliary list called the Active List, that contains indices of queues that have at least one packet and to avoid examining empty queues. Then, selects a non-empty queue and the deficit counter of that queue is incremented by its quantum value that makes the deficit counter the maximum number of bytes that can be sent at each turn. Next, we check if the value of deficit counter is greater than packet size at the head of the queue if so, we can send that packet and decrement the deficit counter value by the packet size. For the next turn, the size of the next packet is compared to the new counter value and so on to the end of the queue. Once the queue is empty the scheduler will skip to the next queue and remove the queue from the active list and set the value of the deficit counter to 0. Bear in mind, for the active list when an empty queue arrives, a packet and index of that queue added to the list. IV. SIMULATION RESULTS AND DISCUSSION As a simulation tool, we used OPNET (Optimized Network Engineering Tool), OPNET is extensive and powerful simulation tool environment for building, performance

62

2018 9th International Conference on Information and Communication Systems (ICICS)

analysis for different types of networks with various protocols [17]. Under Windows environment, we used OPNET simulator version 17.5 with 802.11n 5.0GHZ (high throughput) as a maximum limitation of technology used, 65 (base)/ 600 Mbps as maximum data rate and 1 packet/sec as packet sending rate. This scenario measures the performance of cloud network that contains four subnets (China, London, Malaysia, and Cameroon), one router and one cloud. China and Malaysia subnets are the same and contain ten Ethernet workstations connected to one Ethernet switch, two routers called (Cisco C3600 Router) and one firewall. Cameroon and London subnets are also the same, with seven Ethernet workstations, four Ethernet servers connected to Ethernet switch, two routers called (Cisco C3600 Router) and one firewall. The main router in each subnet connected to the router (Cisco C4000 Router) outside subnets, where this router is connected directly to subnets and to IP cloud by PPP_DS1 link, all Ethernet workstations are connected to the switches by bi-directional 10Base_T links. Network performance is analyzed for four applications: database server with a heavy load, Email server with a heavy load, FTP server with a heavy load and the HTTP server with a heavy load. Fig. 2 shows network model for this scenario, simulation time is 1 hour. The results are shown and discussed in the next section.

Fig. 3.1, Shows Ethernet average delay. When network has enough load, MDRR start to achieve better average delay. On average, MDRR achieve 0.005848 and FIFO achieve in 0.0065 for minimum average delay. Thus, in this scenario MDRR achieve better delay.

Fig. 3.1: Ethernet Average delay (sec) for FIFO and MDRR.

2. Packet Delivery Ratio in Database Query Fig. 3.2, shows the average traffic sent and received in database query. MDRR achieve 16.839 (packet/sec) compared with FIFO with 9.637 (packet/sec), MDRR achieve 5.613 (packet/sec) compared with FIFO with 3.212 (packet/sec) as maximum average traffic received value. As a result, MDRR has better average traffic sent and received over FIFO in database query.

Fig. 2: Network Structure.

Application configuration node, profile configuration node, and QoS attribute node are used. Application configuration node is used to configure all network applications for all services needed by clients, profile configuration node used to define client profiles, where each profile can contain one or more applications, and QoS attribute node used to define queuing methods that network will run according to. A. Simulation Results for Overall system 1.

Fig. 3.2: Average traffic sent and received in Database Query.

3. Packet Delivery Ratio in Email Fig. 3.3, shows the average traffic sent and received (packet/sec) in Email. MDRR achieve 4.1667 (packet/sec) compared to FIFO with 2.3889 packet/sec as maximum average traffic sent and received values. Thus, MDRR average traffic sent and received is better than FIFO in Email.

Average Ethernet Delay

63

2018 9th International Conference on Information and Communication Systems (ICICS)

1.

Point-to-Point queuing delay.

Fig. 4.1, shows average queuing delay (sec). In this case, MDRR achieves better delay compared to FIFO. MDRR achieve 0.000518 (sec) where FIFO achieve 0.000539 (sec) minimum value of queuing delay. 2.

Point-to-Point Throughput.

Fig. 4.2 shows average point-to-point throughput (packet/sec) for both in-going and out-going for MDRR and FIFO, MDRR throughput is better than FIFO.

Fig. 3.3: Average traffic sent and received (Packet/sec) in Email.

4. Packet Delivery Ratio in FTP Fig. 3.4, shows the average traffic sent and received (packet/sec) in FTP. MDRR achieve 1.2778 (packet/sec) compared with FIFO that achieve 0.9167 (packet/sec) as maximum average traffic sent value, and MDRR achieves 0.5278 (packet/sec) compared to FIFO of 0.3333 (packet/sec) maximum average traffic received value. MDRR achieve better average traffic sent and received over FIFO in FTP.

Fig. 4.2: Average Point-to-Point Throughput.

3.

Average Point-to-Point Utilization

Fig. 4.3, when throughput (data transfer rate) is better, utilization (the ratio of current network traffic to the maximum traffic that the port can handle) will be better too. So MDRR achieve better throughput, it will achieve better utilization too than FIFO.

Fig. 3.4: Average traffic sent and received in FTP.

B. Between Malysia and Cisco router 4000:

Fig. 4.3: Average Point-to-Point Utilization.

C. Between Cameroon and Cisco router 4000: 1.

Fig. 4.1: Average Point-to-Point Delay.

Point-to-Point queuing delay.

Fig. 5.1, shows average queuing delay (sec). In this case MDRR achieve better delay than FIFO. MDRR achieve 0.001302 (sec) where FIFO achieve 0.001447 (sec) as minimum value of queuing delay.

64

2018 9th International Conference on Information and Communication Systems (ICICS)

V.

CONCLUSION

Both MDRR and FIFO are used queuing method in 5G. When MDRR is used instead of FIFO queuing method, the network performance is enhanced, and network achieves better queuing delay, better throughput, better utilization, better average sent/ received packet delay and total power consumption will be minimized. It was proved that The MDRR is better than FIFO in many metrics at different scenarios, such as traffic packet dropped delay, end-to-end delay for Video and VoIP applications, average Ethernet delay (sec), and average delay (sec) in WLAN, and many other. REFERENCES Fig. 5.1: Average Point-to-Point Delay.

2.

Point-to-Point Throughput

Fig. 5.2, shows average point-to-point throughput (packet/sec) for both in-going and out-going for MDRR and FIFO, MDRR achieve better throughput than FIFO.

[1] [2] [3] [4]

[5] [6]

[7]

[8]

Fig. 5.2: Average Point-to-Point Throughput.

3.

Point-to-Point Utilization

Fig. 5.3, when throughput (data transfer rate) is better, utilization (the ratio of current network traffic to the maximum traffic that the port can handle) will be better too. So MDRR achieve better throughput, it will achieve better utilization too than FIFO.

[9] [10] [11] [12] [13] [14]

[15]

[16] [17]

Warren, D., and C. Dewar. "Understanding 5G: Perspectives on future technological advancements in mobile." GSMA Intelligence Report (2014). Dass, Deepak Kumar, Parvind Kumar, and B. Tech Student. "Modern Wireless 5G Technology: Future of Super Fast Internet." International Journal of Engineering Science 10253 (2017). Hong, Xuemin, et al. "Cognitive radio in 5G: a perspective on energy-spectral efficiency trade-off." IEEE Communications Magazine 52.7 (2014): 46-53. Mavromoustakis, Constandinos X., et al. "An energy-aware scheme for efficient spectrum utilization in a 5G mobile cognitive radio network architecture." Telecommunication Systems 59.1 (2015): 63-75. Michele Garetto, Don Towsley (2003) 'Modeling, Simulation and Measurements of Queuing Delay under Long-tail Internet Traffic', San Diego, California, USA. , (), pp. Mustafa, Mustafa EG, and Samani A. Talab. "The Effect of Queuing Mechanisms First in First out (FIFO), Priority Queuing (PQ) and Weighted Fair Queuing (WFQ) on Network’s Routers and Applications." (2016). Mohammed Syed Dar, Pradeep Sharma, Priyanka Tiwari (2016) 'Efficient Scheduling For D2D Communication in 5g Networks', International Research Journal of Engineering and Technology (IRJET) , Volume: 03 (Issue: 05 ), pp. Jafari, Fahimeh, Axel Jantsch, and Zhonghai Lu. "Weighted Round Robin Configuration for Worst-Case Delay Optimization in Network-on-Chip." IEEE Transactions on Very Large Scale Integration (VLSI) Systems 24.12 (2016): 3387-3400. J. Y. L. Boudec and P. Thiran, ”Network Calculus: A Theory of Deterministic Queuing Systems for the Internet”, Number 2050 in LNCS, 2004. Bousia, A. (2014). Sharing the small cells for energy efficient networking: How much does it cost? In IEEE GLOBECOM, Austin, TX. Antonopoulos, A., & Verikoukis, C. (2014). Multi-player game theoretic MAC strategies for energy efficient data dissemination. IEEE Transactions on Wireless Communications, 13(2), 592–603. Ericsson. (2013). Technology for Good: Ericsson Sustainability and Corporate Responsibility Report 2012. Stockholm: Ericsson. Retrieved March 11, 2014. Singh, S., & Singh, P. (2012). Key concepts and network architecture for 5G mobile technology. International Journal of Scientific Research Engineering and Technology, 1(5), 165–170. Li, Bin, Atilla Eryilmaz, and R. Srikant. "Emulating Round-Robin in Wireless Networks." Proceedings of the 18th ACM International Symposium on Mobile Ad Hoc Networking and Computing. ACM, 2017. Olsson, M., Cavdar, C., Frenger, P., Tombaz, S., Sabella, D., & Jantti,R.(2013).5GrEEn:TowardsGreen5Gmobilenetworks.In Proceedings of the IEEE 9th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob) (pp. 212–216). Rowell, C., Han, S., Xu, Z., Li, G., & Pan, Z. (2014). Toward green and soft: A 5G perspective. IEEE Communications Magazine, 52(2), 66–73. OPNET Technologies, Inc. http://www.opnet.com.

Fig. 5.3: Average Point-to-Point Utilization.

65