2 VirtualClock Algorithm3 VirtualClock Implementation5 Simulation ExperimentsAs seen in Figures 7 and 11 (bottom), Client 3 generates self-similar traffic. The Virtual Clock algorithm forwards packets at rate of 4 packets/sec, which is the average traffic generation rate of the source.AcknowledgmentReferencesImplementation of VirtualClock Scheduling Algorithm in OPNETNazy Alborz and Ljiljana TrajkovićSchool of Engineering ScienceSimon Fraser UniversityVancouver, British ColumbiaCanada V5A 1S6{nalborz, ljilja}@cs.sfu.cahttp://www.ensc.sfu.ca/research/cnl AbstractIn today’s high-speed packet networks that support variousapplications with different service requirements, congestioncontrol is an important issue. One of the methods forpreventing congestion is packet scheduling. Packet schedulingin routers can provide guaranteed performance in terms ofdelay, delay jitter, packet loss, and throughput. In this paper, we describe the OPNET model of an IP routerwith a scheduling algorithm called VirtualClock. TheVirtualClock algorithm monitors the average transmissionrates of packet data flows. It also provides each flow with aguaranteed throughput and a low queuing delay. We haveincorporated the VirtualClock algorithm into the OPNETprocess model ip_output_iface. This process model executesscheduling algorithms in the network layer of IP objects. 1 IntroductionOne of the most significant promises in today’s integratedservices packet-switched networks (such as Internet) isproviding Quality of Service (QoS) guarantees [1]. Thesenetworks are able to integrate applications with a wide rangeof traffic characteristics. These applications range from videoconferencing with stringent QoS requirements, to best effortapplications with no required guarantees. In network switchingnodes, packets with different QoS requirements interact withone another when they are multiplexed at the same outputport. If there is no control over these interactions, they willdegrade the performance of the network [2].Packet scheduling algorithms in network routers and switchescan provide guaranteed QoS. Scheduling algorithms not onlyallow packets from various traffic streams to be statisticallymultiplexed, but also provide a protection between thesestreams. The three main functions of packet schedulingdisciplines are to determine:- which packets among different service classes gettransmitted- when these packets get transmitted, and - which packets get discarded in case of an overflow inswitch buffers.-These operations affect throughput, delay, and loss rateperformance parameters. Packet scheduling mechanisms are classified into twocategories: work-conserving and non-work-conserving [3]. Awork-conserving scheduler is idle when there are no packetswaiting in the router’s queues. In a non-work-conservingscheduler, each packet is assigned a time when it has to be sentto the output interface. The scheduler remains idle and nopacket will be transmitted until the time when the next packetis eligible.In this paper we describe the implementation of a work-conserving scheduling algorithm called VirtualClock [4]. Thealgorithm is similar to the Weighted Fair Queuing (WFQ)algorithm [1] that is currently implemented in OPNET [5].The difference between the WFQ and VirtualClock algorithmsis that the VirtualClock simplifies the calculation of finishtimes. Finish time is calculated and used by the schedulerwhen choosing the next packet to be dequeued [6]. Otheradvantages of the VirtualClock algorithm are:- it provides per connection bandwidth allocation- it guarantees protection between traffic flows.This paper is organized as follows. In Section 2, we introducethe VirtualClock [4] scheduling algorithm, its role, and itsfunctionality. In Section 3, we describe the implementation ofthe algorithm in the IP layer process module of an IP-routernode model. The validity of the model through OPNETsimulations of a simple network is discussed in Section 4.Section 5 presents the simulation results from a network withvarious traffic sources, which employs VirtualClockalgorithm. We conclude with Section 6. 2 VirtualClock Algorithm The idea behind the VirtualClock algorithm was derived fromTime Division Multiplexing (TDM) systems. A TDM systemeliminates interference among users because individual userchannels (flows) can transmit only during specific time slots.The disadvantage of a TDM system is that users are limited toconstant data transmission rates and the channel capacity iswasted whenever a slot is given to a flow that has no data tosend at that moment. The purpose of the VirtualClock1algorithm is to maintain the guaranteed throughput andfirewalls of a TDM system, while still achieving the statisticalmultiplexing properties of packet switched networks. The algorithm makes the statistical data flow resemble a TDMchannel by assigning each data flow a “virtual clock.” Each“virtual clock” advances one tick at every packet arrival froma specific flow. The tick step is the mean packet inter-arrivaltime that has been specified by the flow. Thus, each “virtualclock” carries the expected arrival time of the packet. If a flowsends packets according to its specified average rate, its“virtual clock” follows real time. The algorithm stamps thepackets with their own “virtual clock” values and transmits thepackets in ascending order of these stamps. Nevertheless, thereis a major difference between a TDM system and a networkcontrolled by a VirtualClock scheduling algorithm. Thedifference is that unlike TDM, a VirtualClock controllednetwork can support data flows with distinct throughput rates.The network reservation protocol determines how large ashare of bandwidth each flow needs on average. Then,according to the flow’s reserved transmission rate, theVirtualClock algorithm determines which packet should beforwarded next in the case that there is more than one packetwaiting [4].We consider the