• 沒有找到結果。

Design of a power-spectrum-based ATM connection admission controller for multimedia communications

N/A
N/A
Protected

Academic year: 2021

Share "Design of a power-spectrum-based ATM connection admission controller for multimedia communications"

Copied!
8
0
0

加載中.... (立即查看全文)

全文

(1)

troller contains a power-spectrum-indexed table for managing multimedia call requests, where traffic characteristics of call re-quests are described by the power spectrum. The power spectrum can be obtained from the claimed traffic parameters of peak rate, mean rate, and peak rate duration; the power spectrum has been shown to have a dominant effect on system performance. The results show that the proposed power-spectrum-based connec-tion admission control method achieves higher system utilizaconnec-tion and lower call-blocking probability than the equivalent-capacity allocation method.

Index Terms—Asynchronous transfer mode, connection

admis-sion control, multimedia communications, power spectrum.

I. INTRODUCTION

A

SYNCHRONOUS transfer mode (ATM) is considered to be the foundation on which broad-band Integrated Services Digital Networks (B-ISDN) are built [1], [2]. ATM systems have the advantage of flexibility and efficiency in accommodating multimedia services that possess different traf-fic characteristics and quality-of-service (QoS) requirements. However, ATM switches require a sophisticated connection admission control method, so as not only to satisfy distinct QoS requirements, but also to enhance system utilization.

An equivalent-capacity allocation method has been pro-posed [3]. These studies on resource allocation and manage-ment for ATM switches assumed that voice sources were mod-eled by a two-state Markov-modulated deterministic process (MMDP), video sources were modeled by a multiple-state birth–death Markov-modulated Poisson process (MMPP), and data sources were modeled by a Poisson process [4]–[6]. Manuscript received October 13, 1996; revised July 20, 1997. This work was supported in part by the National Science Council, Taiwan, R.O.C., under Contract NSC 84-2213-E009-137.

C.-J. Chang and R.-G. Cheng are with the Department of Communication Engineering and Center for Telecommunications Research, National Chiao Tung University, Hsinchu, 300 Taiwan, R.O.C.

C.-H. Lin was with the Department of Communication Engineering and Center for Telecommunications Research, National Chiao Tung University, Hsinchu, 300 Taiwan, R.O.C. He is now with the Switching Technology Department, Chunghwa Telecommunication Laboratories, Chungli, Taiwan, R.O.C.

D.-S. Guan was with the Department of Communication Engineering and Center for Telecommunications Research, National Chiao Tung University, Hsinchu, 300 Taiwan, R.O.C. He is now with the Network Planning Depart-ment, Chunghwa Telecommunication Laboratories, Chungli, Taiwan, R.O.C.

Publisher Item Identifier S 0278-0046(98)00410-9.

blocking probability and throughput) of ATM switches. There-fore, in designing an ATM connection admission controller for multimedia communication, it is necessary to thoroughly consider this correlation function.

Li et al. [7]–[10] studied a high-speed network from the frequency domain of input traffic. They found that the corre-lation function can be captured by second-order and higher order statistics, such as power spectrum, bispectrum, and trispectrum, and the second-order power spectrum has a much more significant effect on the system performance than any high-order statistics.

In this paper, we designed an ATM connection admis-sion controller for multimedia calls (voice, video, and data) based on the power spectra of the cell arrival process of calls. We constructed a connection admission control table that contained the performance measure of cell loss prob-ability versus parameters of bell-shaped power spectrum of voice/video (real-time) calls and arrival rates of data (nonreal-time) calls. When a new real-time call comes to an ATM switch, the connection admission controller in the switch first estimates the power spectrum of the new incoming call from the negotiated parameters of peak rate, mean rate, and peak rate duration, adds the power spectrum of the new call to the power spectrum of existing calls, and then obtains a bell-shaped power spectrum. With the parameters of the approximated bell-shaped power spectrum, the connection admission controller obtains the cell loss probability of the new real-time call by looking up the connection admission control table and then accepts or rejects the new real-time call. Similarly, when a new nonreal-time call comes to the switch, the connection admission controller estimates the traffic load of the existing calls and then judges whether to accept or reject the new nonreal-time call according to its cell loss probability. The results show that our power-spectrum-based connection admission control method achieves higher system utilization and lower call-blocking probabilities than the equivalent-bandwidth allocation method proposed by Gu´erin, Ahmadi, and Naghshineh [3].

This paper is organized as follows. In Section II, we depict the system diagram of an ATM switch and describe the power spectrum of Markov-modulated input processes for multimedia services. In Section III, we describe our design for the power-spectrum-based connection admission controller, including the 0278–0046/98$10.00  1998 IEEE

(2)

Fig. 1. The conceptual diagram of an ATM switch.

establishment of the connection admission control table, for multimedia communications. The simulation results and dis-cussions are given in Section IV. The concluding remarks are presented in Section V.

II. SYSTEMCONFIGURATION ANDINPUTPOWER-SPECTRUM The conceptual diagram of an ATM switch containing the power-spectrum-based connection admission controller is shown in Fig. 1. We classify the multimedia calls input to the ATM switch into real-time voice/video (type-1) and nonreal-time data (type-2) traffic. Input messages from both types of traffic are segmented into fixed-length ATM cells and the constant service time of an ATM cell is referred as a frame time period The switch provides two buffers with capacities and for type-1 and type-2 traffic, respectively, and it has parallel synchronous links (servers) connected to the switching network. The parallel synchronous links are operated in a slot-by-slot fashion. A queue-length-threshold (QLT) link capacity assignment policy [11], [12] is adopted, which is noted as follows. If the queue length of type-2 traffic is higher than or equal to a threshold value links are reserved for type-2 traffic and the remaining links are first accessed by type-1 traffic and then accessed by type-2 traffic; if the queue length of type-2 traffic is less than the total links are first allocated to type-1 traffic and type-2 traffic can only utilize the remaining links. The service discipline in each buffer is first-come-first-serve (FCFS).

We use a two-state MMDP to model the cell-arrival process of an incoming voice call. As shown in Fig. 2(a), and 0 are the cell-arrival rates when the voice source is at “on” state and “off” state, respectively; and are the transition rates for the two-state Markov chain. We model the cell-arrival process of an incoming video call by a superposition of independent and identical two-state MMPP’s with cell-arrival rate at “on” state, the cell-arrival rate 0 at “off” state, and transition rates of and Fig. 2(b) shows the video source model for an -state MMPP, and we use an independent Poisson arrival process with a mean rate of for the cell-arrival process of an incoming data call. All the voice, video, and

(a)

(b)

Fig. 2. (a) Voice source: a two-state MMDP. (b) Video source:(N +1)-state MMPP.

data call-arrival processes are assumed to be independently Poisson distributed.

Denote as the transition-rate matrix of an MMPP input video traffic to the switch, and can be represented as

(1)

where is the eigenvalue of and are the associated right and left eigenvectors of respectively, and

is the number of states in the MMPP [7]. The autocorrelation function and its corresponding power spectrum of the input process can be obtained by [7]

(2) and

(3)

(6) is the average power contributed by and is given by

for (7)

and are the th and th entities of the vectors and respectively; is the bell function given by

for (8)

and

(9)

is defined as the half-power bandwidth of a bell-shaped function corresponding to the eigenvalue and the derivation

of can be found in Appendix A; is the

central frequency of the bell-shaped function defined as (10) Equation (3) shows that the transition-rate matrix of an MMPP has a power spectrum consisting of a white noise a dc component and a set of bell-shaped functions the position of which is centered at

The white noise results from from the Poisson local dynamics of the MMPP.

As far as the MMDP voice traffic is concerned, the Markov-ian property of the Markov chain still generates the bell-shaped functions and the dc component, but the deterministic rate process in the state of the Markov chain generates another dc component, instead of white noise.

Li and Huang in [7] showed that the influence of the white noise on the queueing behavior of a system was negligible. Therefore, in this paper, we only considered the dc component

the average power and the half-power bandwidth of the bell-shaped functions as parameters for the power-spectrum-based connection admission control method.

and

(13) When a new video-call request with arrives, we can also obtain the arrival rate and the transition rates and

of the -state MMPP, as shown in Fig. 2(b), by (14) (15) and

(16)

The detailed derivation of the transition rate is given in Appendix B.

We then have the corresponding input power spectrum

characterized by parameters and as

given in (6)–(8). Note that, for the -state MMPP video-source model contributed by independently identically distributed (i.i.d.) two-state Markov chains,

and the bell-shaped functions are all zero centered.

III. POWER-SPECTRUM-BASED CONNECTION ADMISSION CONTROL

We describe the design of the proposed power-spectrum-based connection admission controller and the construction of the power-spectrum-based connection admission control table here. A new call is assumed to be accepted if the ATM switch can satisfy QoS requirements of cell loss probability for new and existing calls. If the requirement cannot be satisfied, the incoming call is rejected.

A. Power-Spectrum-Based Connection Admission Controller The functional block diagram of the power-spectrum-based connection admission controller is shown in Fig. 3. Whenever a new voice or video real-time call comes to the switch, the power spectrum estimator computes the arrival rate and the transition rates of the Markov modulated arrival process (MMDP or MMPP) and estimates the power spectrum from the

(4)

Fig. 3. The functional block diagram of the power-spectrum-based connec-tion admission controller in an ATM switch.

Fig. 4. An example of the approximated bell-shaped function obtained from a composite power spectrum.

negotiated traffic parameters The spectrum com-poser adds the power spectrum of the new call to the power spectrum of existing calls and obtains a composite power spectrum. This composite power spectrum is approximated by a bell-shaped function. (Fig. 4 shows an example of an approximated bell-shaped function representing the composite power spectrum obtained from the power spectrum of 100 existing voice calls in progress and one new video call in call setup.) Consequently, the cell-loss probability of the new real-time call can then be found by looking up the connection admission control table. (The establishment of the power-spectrum-based connection admission control table will be described later.) The admission decision maker accepts the new call if cell-loss probability satisfies its QoS requirement, otherwise it rejects the new call.

When a new nonreal-time data call requests to enter the switch, the traffic load estimator obtains the traffic load in-tensity of the new call according to its mean arrival rate The total traffic load estimator adds the load of the new call to the load of existing nonreal-time calls and obtains the overall traffic intensity The cell-loss probability of the type-2 traffic can be found by looking it up in the the connection admission control table, and the admission decision maker accepts the new call if the cell-loss probability satisfies

its QoS requirement, otherwise it rejects the new call. Here, we elaborate on how to approximate the composite power spectrum by a bell-shaped function in the spectrum composer. As (8) implies, a zero-centered bell-shaped function can be characterized by its average power and half-power bandwidth. We let the approximated bell-shaped function of the composite power spectrum possess the same value at and the same average power as those of the composite power spectrum. We denote the dc component and the approximated bell-shaped function of the composite power spectrum by and and denote the average power and the half-power

bandwidth of by and The power spectrum of

existing calls is assumed to have a dc component and an approximated bell-shaped function with average power and half-power bandwidth the power spectrum of a new call obtained from the negotiation parameters

also has a dc component and an approximated bell-shaped function with average power and half-power

bandwidth Therefore, and can be

respectively obtained by

(17) (18) and

(19)

Equation (19) is derived from at

Such an assumption would cause the approximated bell-shaped function to contain more low-frequency component power and result in an overestimation of type-1 cell-loss probability in the construction of the connection admission control table. B. Establishment of the Power-Spectrum-Based Connection Admission Control Table

The power-spectrum-based connection admission control table contains cell-loss probabilities of type 1 and type 2 (CLP1 and CLP2) versus and of zero-centered bell-shaped functions of type-1 traffic and of type-2 traffic. We use simulations to obtain CLP1 and CLP2. The simulation model for obtaining CLP1 for a given and is the same as that defined in Fig. 1, which adopts QLT policy and has system parameters of and The input process of type-2 data traffic is assumed to be a Poisson process with an arrival rate of The input process of type-1 voice (or video) traffic is assumed to be two-state -state) Markov chains with arrival rates of (or and transition rates of and (or and obtained

from parameters of and of a zero-centered

bell-shaped function given by

or (20)

or (21)

and

(5)

FUNCTION cac table BEGINFUNC

Initialize the values of traffic load for type-2 traffic and dc component ,

average power , and half-power bandwidth for type-1 traffic

Set the desired ranges of , and

Set the increments steps of ,

and

For within the desired range of traffic load Call type-2 simulation

For within the desired range of bandwidth For within the desired range of average

power

For within the desired range of dc component

Call power-spectrum to 2-state-MC Call type-1 simulation

Write to cac table with format , CLP2, CLP1) Increase by EndFor Increase by EndFor Increase by EndFor Increase by EndFor ENDFUNC

FUNCTION type-2 simulation BEGINFUNC

Generate type-2 traffic according to a Poisson process with mean rate

Simulate the system described in the second paragraph of this subsection that has buffer size,

available links, and queue-length-threshold. Calculate the CLP2

Return CLP2 ENDFUNC

Generate type-2 input traffic according to the Poisson process with mean rate

Simulate the system described in the first paragraph of this subsection that adopts the QLT scheduling policy and has system parameters of

, and . Calculate CLP1 Return CLP1 ENDFUNC

IV. SIMULATION EXAMPLES AND DISCUSSIONS In the simulations, the system parameters were given by the buffer sizes the total number of links the number of reserved links and the queue-length threshold of the QLT policy For simplicity, the QoS requirement of cell-loss probability was assumed to be for voice service, for video service, and for data service. We considered several simulation examples, but showed only an example in which a switch supports multimedia services of voice, video, and data calls.

In the example, the call parameters were defined as follows: call arrival rates for voice, video, and data were 10.115, 0.0743, and 0.4246 calls/s; the call holding time for voice, video, and data were 1, 5, and 30 s; and the peak rate was 64 kb/s, the mean rate was 27.6 kb/s, and the peak duration was 1.366 s for a voice call. was 6 Mb/s, was 2.04 Mb/s, and was 3.86 s for a video call. The arrival rate for a data call was 9.0 cells/slot. We compared results with the system using the equivalent-capacity allocation method proposed in [3], under the same call parameters mentioned above.

Fig. 5(a) shows the number of accepted voice calls versus time. In it, we can see that the number of accepted voice calls using both control methods is similar. Fig. 5(b) shows the voice-call-blocking probability versus time. It shows that the power-spectrum-based control method outperforms the equivalent-capacity allocation method. Fig. 5(c) shows that the number of accepted data calls for both control meth-ods is also similar. Fig. 5(d) shows the number of accepted video calls versus time. We can see that the system using the power-spectrum-based control method can accommodate more video calls than the one using the equivalent-capacity

(6)

(a)

(b)

(c)

Fig. 5. (a) Number of accepted voice calls versus time. (b) Voice-call-blocking probabilities versus time. (c) Number of accepted data calls versus time.

allocation method. Fig. 5(e) shows the video-call-blocking probability versus time. It also reveals that the system us-ing the power-spectrum-based control method has a lower video-call-blocking probability than the one using equivalent-capacity allocation method, and Fig. 5(f) shows overall system utilization versus time. In the simulations, the cell-loss prob-abilities are all zero using both methods. It shows that the power-spectrum-based connection admission control method can achieve 9% greater system utilization than the equivalent-bandwidth allocation method. Therefore, the power-spectrum-based connection admission control method is more efficient than the equivalent-capacity allocation method, because the former method takes the correlation (second-order statistics) effect of cell-arrival process into account.

V. CONCLUDING REMARKS

This paper has designed a power-spectrum-based connec-tion admission controller for ATM switches in multimedia communication applications. Such a power-spectrum-based connection admission controller in the switch takes the cor-relation characteristics (second-order statistics) of cell-arrival

processes into account. Whether an incoming call is accepted by a switch is determined by whether the cell-loss probability, obtained by reference to a power-spectrum-based connection admission control table, satisfies the QoS requirement or not. We presented a procedure for establishing the power-spectrum-based connection admission control table which contains the cell-loss probability indexed by type-1 traffic power-spectrum parameters and type-2 traffic arrival rate. In order to sim-plify the lookup table, we approximated the composite power spectrum, which is the sum of the power spectra of existing calls and new incoming calls, as a bell-shaped function. The results show that the system using the power-spectrum-based connection admission control method achieves a 9% higher utilization than the one using the equivalent-capacity allocation method [3].

In a real ATM network, where many ATM nodes are included, the CAC should be made for the global decision, instead of the local decision studied in this paper. Because the acceptance or rejection of the global decision is determined by the local decision of each node traversed by the call, the scheme we proposed can be extended to the global decision.

(7)

(e)

(f)

Fig. 5. (Continued.) (d) Number of accepted video calls versus time. (e) The video-call-blocking probabilities versus time. (f) The overall system utilization versus time, for a switch that accepts voice, data, and video calls.

APPENDIX A

Consider a bell-shaped power spectrum with a half-power bandwidth From (8) and (9), must satisfy

(A1) We can then find from (A1), which is given by

(A2)

APPENDIX B

We know that the steady-state probability of the -state MMPP constructed by i.i.d. two-state Markov chain is

binominally distributed. The probability that there are

two-state Markov chains in the “on” two-state, is

given by

(B1) where is defined as

(B2) Therefore, we can have the mean rate of the -state MMPP by

(B3) Using (14), (15), and (B3), we can obtain the transition rate

by

(8)

REFERENCES

[1] R. H¨andel and M. M. Huber, Integrated Broadband Networks: An Introduction to ATM-Based Networks. Reading, MA: Addison-Wesley, 1991, p. 14.

[2] “Traffic control and congestion control in B-ISDN,” International Telecommunications Union-Telecommunications, Geneva, Switzerland, ITU-T: Recommendation I.371, May 1996.

[3] R. Gu´erin, H. Ahmadi, and M. Naghshineh, “Equivalent capacity and its application to bandwidth allocation in high-speed networks,” IEEE J. Select. Areas Commun., vol. 9, pp. 968–981, Sept. 1991.

[4] J. N. Daigle and J. D. Langford, “Models for analysis of packet voice communications systems,” IEEE J. Select. Areas Commun., vol. SAC-4, pp. 856–868, Sept. 1986.

[5] H. Heffes and D. M. Lucantoni, “A Markov modulated characterization of packetized voice and data traffic and related statistical multiplexer performance,” IEEE J. Select. Areas Commun., vol. SAC-4, pp. 856–868, Sept. 1986.

[6] B. Maglaris , D. Anastassiou, P. Sen, G. Karlsson, and J. Robbins, “Per-formance analysis of statistical multiplexing for packet video sources,” IEEE Trans. Commun., vol. 36, pp. 834–844, July 1988.

[7] S. Q. Li and C. L. Hwang, “Queue response to input correlation functions: continuous spectral analysis,” IEEE Trans. Commun., vol. 41, pp. 678–692, Dec. 1993.

[8] , “Queue response to input correlation functions: discrete spectral analysis,” IEEE/ACM Trans. Networking, vol. 1, pp. 522–533, Oct. 1993. [9] S. Q. Li, S. Chong, and C. L. Hwang, “Link capacity allocation and network control by filtered input rate in high speed networks,” IEEE/ACM Trans. Networking, vol. 3, pp. 10–25, Feb. 1995. [10] H. D. Sheng and S. Q. Li, “Second order effect of binary sources on

characteristics of queue and loss rate,” in Proc. IEEE INFOCOM, Apr. 1993, pp. 18–27.

[11] C. J. Chang, P. C. Lin, and J. M. Chen, “Study on optimal queue-length-threshold scheduling policy for an ATM multiplexer with finite buffers and batch Poisson arrivals,” Comput. Networks ISDN Syst., no. 26, pp. 524–540, 1994.

[12] R. Chipalkatti, J. F. Kurose, and D. Towsley, “Scheduling policies for real-time and nonreal-time traffic in a statistical multiplexer,” in Proc. IEEE INFOCOM, 1989, pp. 774–783.

Chung-Ju Chang (S’81–M’85–SM’94) was born

in Taiwan, R.O.C., in 1950. He received the B.E. and M.E. degrees in electronics engineering from National Chiao Tung University, Hsinchu, Taiwan, R.O.C., in 1972 and the Ph.D degree in electrical engineering from National Taiwan University, Tai-wan, R.O.C., in 1985.

From 1976 to 1988, he was with Telecommunica-tion Laboratories, Directorate General of Telecom-munications, Ministry of ComTelecom-munications, Taiwan, R.O.C., as a Design Engineer, Supervisor, Project Manager, and then Division Director. He was involved in designing digital switching system, RAX trunk tester, ISDN user–network interface, and ISDN service and technology trials in the Science-Based Industrial Park. He also acted as a Science and Technical Advisor for the Minister of the Ministry of Communications from 1987 to 1989. In August 1988, he joined the faculty of the Department of Communication Engineering and Center for Telecommunications Research, College of Electrical Engineering and Computer Science, National Chiao Tung University, Hsinchu, Taiwan, R.O.C., where he is currently a Professor and also an Advisor for the Ministry of Education. He was Director of the Institute of Communication Engineering from August 1993 to July 1995. His research interests include performance evaluation, asynchronous transfer mode (ATM) networks, and personal communication service (PCS) networks.

Dr. Chang is a member of the Chinese Institute of Engineers (CIE).

Chih-Hen Lin received the B.E. and the M.E.

degrees in communication engineering from National Chiao Tung University, Hsinchu, Taiwan, R.O.C., in 1992 and 1994, respectively.

Since 1994, he has been an Assistant Design Engineer in the Switching Technology Depart-ment, Chunghwa Telecommunication Laboratories, Chungli, Taiwan, R.O.C., where he was involved in constructing and testing the ATM broad-band network.

Da-Sheng Guan received the B.E. and the M.E. degrees in communication engineering from National Chiao Tung University, Hsinchu, Taiwan, R.O.C., in 1992 and 1994, respectively.

Since 1994, he has been an Assistant De-sign Engineer in the Network Planning Depart-ment, Chunghwa Telecommunication Laboratories, Chungli, Taiwan, R.O.C., where he was involved in designing the call control procedure of ISDN PABX. He is currently involved in the software design of the Layer 3 protocol of the wireless subscriber module (WSM) in the personal communications test platform (PCTP) system and the call control procedure of the PCTP system.

Ray-Guang Cheng (S’94–M’97) received the B.E.,

M.E., and Ph.D. degrees from National Chiao Tung University, Hsinchu, Taiwan, R.O.C., in 1991, 1993, and 1996, respectively, all in communication engi-neering.

His current research interests include performance analysis, ATM networks, fuzzy systems, and neural networks.

數據

Fig. 2. (a) Voice source: a two-state MMDP. (b) Video source: (N +1)-state MMPP.
Fig. 3. The functional block diagram of the power-spectrum-based connec- connec-tion admission controller in an ATM switch.
Fig. 5. (a) Number of accepted voice calls versus time. (b) Voice-call-blocking probabilities versus time
Fig. 5. (Continued.) (d) Number of accepted video calls versus time. (e) The video-call-blocking probabilities versus time

參考文獻

相關文件

(ii) Maximum power point tracking (MPPT) controller design of the PV module.. (iii) MPPT controller design of the WTG without sensing the

(d) While essential learning is provided in the core subjects of Chinese Language, English Language, Mathematics and Liberal Studies, a wide spectrum of elective subjects and COS

For ASTROD-GW arm length of 260 Gm (1.73 AU) the weak-light phase locking requirement is for 100 fW laser light to lock with an onboard laser oscillator. • Weak-light phase

The temperature angular power spectrum of the primary CMB from Planck, showing a precise measurement of seven acoustic peaks, that are well fit by a simple six-parameter

PEERS for young adults : Social skills training for adults with autism spectrum disorder and other

Experiment a little with the Hello program. It will say that it has no clue what you mean by ouch. The exact wording of the error message is dependent on the compiler, but it might

For MIMO-OFDM systems, the objective of the existing power control strategies is maximization of the signal to interference and noise ratio (SINR) or minimization of the bit

Table 3 shows the average recall values for all query models using the proposed elevation descriptor (ED), spherical harmonics (SH), 3D shape spectrum descriptor (SSD), and D2.