Adaptive algorithm：

In our algorithm we can clear see the usage of the CPU is as low as round robin. And like the same as round robin, the usage will fall and stand idle till the system has come back to its stable status.

SlaveA CPU usage

0 5 10 15

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 10Sec

% AD 120 calls

AD 140 calls

Figure 31 Adaptive SlaveA CPU SlaveB CPU usage

0 5 10 15

1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 10Sec

% AD 120 calls

AD 140 calls

Figure 32 Adaptive SlaveB CPU

Memory performance：

As mentioned before, the calls consumes the memory instead the usage of the CPU, so not only our algorithm will not effect the performance of the system (Our algorithm works on many scripts, and scripts only increase the usage of the CPU), but also can reduce the average usage of the memory shown in figure 33 and figure 34.

SlaveA mem usage

Figure 33 Adaptive SlaveA Memory

SlaveB mem usage

Figure 34 Adaptive SlaveB Memory

Success and failure number：

At the end, the success and failed calls are much the same as round robin, this is because it is limited by the system’s hardware equipment. We already know that it only consumes the memory, therefore, if wish to handle more calls in the future, the only thing need to be done is add more memory on the board.

1000

Figure 35 Adaptive success and failure numbers 4.4.4 DHS/SHS algorithm：

DHS/SHS is used for there is a big number of end users, and the users are random, the more the users are irregular, the better performance will be shown. Although we have 20 computers as test client, trying to create a random end user environment for hash algorithm, but still, most of the calls are still sent to a single server

In figure 36 we can see when the concurrent calls comes to 140 calls per second, the average use of the CPU is high, this increases the rate of the loose call.

SlaveA CPU Usage

Figure 36 DHS/SHS slaveA CPU

Figure 37 is the performance of the CPU in the slave server B, we can see the usage drops rapidly when the time passes through 160 seconds, this means the system has collapsed, and most of the incoming calls has been refused, therefore cause the usage of the CPU is low in a long period of time.

SlaveB CPU usage

0 5 10 15 20

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 10 Secs

%

Hash 120 calls Hash 140 calls

Figure 37 DHS/SHS slaveB CPU

The rise of the memory usage of concurrent 140 calls is lower than 120 calls, this is because most of the calls has been refused when the concurrent calls is 140, the system is unable to handle so many incoming calls at the same time by using such algorithm, and the reason is end users are not irregular.

Most of incoming calls are sent to slaveB, there for when the concurrent calls reaches to 140 calls per second, the average usage of the memory is lower than slaveA, the result can be compared in figure 38 and figure 39.

SlaveA mem usage

Figure 38 DHS/SHS slaveA memory

SlaveB mem usage

Figure 39 DHS/SHS slaveB memory

As mentioned, the loose call rate rises rapidly when the concurrent calls reaches 140 calls per second

Figure 40 DHS/SHS successful and failure numbers

4.4.5 Algorithms comparison：

Show in figure 41 and 42, we can see the difference more clearly. When the call is under 120 calls, the usage of the cpu is not over 15 percent of the system, and also fully in use. But when the calls increase to 140 calls, the use of the CPU drops and cause the system becoming idle, the reason of this is the same reason we mentioned above, this raises the failure call rate, in another word, this is the sign of showing the system is collapsing.

SlaveA all algorithm CPU comparison

0

Figure 41 All CPU comparison in slaveA

SlaveB all algorithm CPU comparison

0

Figure 42 All CPU comparison in slaveB

Like show in figure 43 and figure 44, we can clearly see the comparison between the average of the memory usage in slaveA and slaveB, the average has shown clearly the

average memory usage of our theme is lesser than all the others. The test is now under only two slave servers, we believe the difference will be more obvious when the slave server are within n servers.

SlaveA all mem algorithm comparison

0

Figure 43 All memory comparison in slaveA SlaveB All mem algorithm comparison

0

Figure 44 All memory comparison in slaveB

Chapter 5 Conclusion and future work

Chapter 5

Conclusion and future work

________________________________________________________

5.1 Conclusion：

Our theme present two stage of the load balance, first stage, using multi server to set up a VOIP cluster system, this system not only can choose the appropriate transportation for the user, but also, providing load balance for the system and keep the system in a stable situation, facing the growth of the user, the system will be able to distribute the load into different servers, and increase system’s extensibility. Second stage, using our adaptive algorithm to distribute the load to a server into a group of slave servers. This way we not only can prevent server breakdown and provide failover support, but also make it easier to manage and extend, cause the changing only need to be done on the master. Furthermore, by using our adaptive algorithm, we not only increase the rate of success calls, but also cause the use of the system’s resource more efficient and equally, at the end, the load shall be distribute to an appropriate server for load balance.

5.2 Future work：

The integration with IPv6：

The use of IPv4 are increasing each day, cause the IP aren’t enough to use, although NAT can settle the problem from now, but this is only a temporally scheme, one day the IP will be totally used. When it comes to this day, the users in the future will be addressed in IPv6, There for, letting IPv4 and IPv6 users run through our system to communicate is a big issue.

Automatically remove memory usage：

Not like in windows environment, most of the data in the memory cash will be deleted after the process is finished, in linux, the data in the memory cash will be saved for a period of time , or until the system is been rebooted. Therefore, if there is a way automatically cleans up the cash memory, the use of the memory will be more efficient and complete.

Reference

[1] A New CPU Availability Prediction Model for Time-Shared Systems Marta Beltra´ n, Antonio Guzma´ n, and Jose Luis Bosque

[2] DYNAMIC LOAD BALANCING FOR DISTRIBUTED NETWORK MANAGEMENT Kiyohito Yoshihara Manabu Isomura Huoki Horiuchi KDDI R&D Loboratories Inc..

2-1-15 Ohara Kamifikuoka-shi Sailam 356-8502, Japan yosshyO kddilabs.jp.

isomuraOkddilabs.jp. hr-horiuchiOkddilabs.jp

[3] Performance Improvement of Cluster System by Server Status Information Chunkyun Youn Department of Internet Software, Honam university [email protected]

[4] LOAD BALANCING IN DISTRIBUTED SHARED MEMORY SYSTEMS An-Chow Lai, Ce-Kuen Shieh, Yih-Tzye Kok, Jyh- Chang Ueng, Ling- Yang Kung. Department of Electrical Engineering, National Cheng Kung University, Tainan, Taiwan.

{ lac,shieh}@eembox.ncku.edu. tw

[5] Suchaya Sanghan, Md Maruf Hasan, “Intelligent P2P VoIP through Extension of Existing Protocols”, The 9th International Conference on Advanced

Communication Technology, February 2007, vol.3, pp.1597 – 1601

[6] D.A Bryan, B.B. Lowekamp, C. Jennings, “SOSIMPLE: A Serverless,

Standards-based, P2P SIP Communication System”, Advanced Architectures and Algorithms for Internet Delivery and Applications, June 2005, pp.42 – 49

[7] J. Fiedler, T. Kupka, T. Magedanz, M. Kleis, “Reliable VoIP Services Using a Peer-to-Peer Intranet”, IEEE International Symposium on Multimedia, December 2006, pp.121 – 130

[8] Kwihoon Kim, Jinsul Kim, Hyun-Woo Lee, Won Ryu, “Method for transmitting SMS for VoIP service supporting Multi-protocol”, IEEE 10th International Symposium on Consumer Electronics, 2006, pp. 1 – 5

[9] V. Paulsamy, S. Chatterjee, “Network convergence and the NAT/Firewall problems”, The 36th Annual Hawaii International Conference on System

Sciences, January 2003, pp. 10

[10] Guo Yihe, Liang Mangui, Guo Yongfang, Zhang Lu, “A design scheme of PSTN media gateway”, The 7th International Conference on Signal Processing, September 2004, vol.3, pp. 2651 – 2654

[11] A Performance Comparison of Dynamic vs. Static Load Balancing Policies in a

Mainframe - Personal Computer Network Mode Hisao Kameda* El-Zoghdy Said Fathyt Inhwan Ryut Jie Lis *ttfUniversity of Tsukuba

[12] IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 9, NO. 3, MARCH 1998 235 A Practical Approach to Dynamic Load Balancing Jerrell Watts and Stephen Taylor

[13] The Application of DLB Algorithm in Distributed Speech Dialogue System

[14] International Telecommunication Union, “Packet based multimedia communications systems,” Recommendation H.323, Telecommunication Standardization Sector of ITU, Geneva, Switzerland, Feb. 1998.

[15] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston, J. Peterson, R. Sparks, M.

Handley, and E. Schooler, “SIP: Session Initiation Protocol,” IETF RFC 3261, June 2002.

[16] S. Berger, H. Schulzrinne, S. Sidiroglou, and X. Wu, “Ubiquitous computing using SIP”, NOSSDAV, pp. 82–89 , 2003.

[17] SIP for Instant Messaging and Presence Leveraging Extensions (SIMPLE) IETF Working Group. http://www.ietf.org/html.

[18] H. Schulzrinne et al. “RTP: A Transport Protocol for Real-Time Applications,” IETF RFC 1889 Jan 1996.

[19] P. Koskelainen, H. Schulzrinne, and X. Wu, “A SIP-based conference control

framework,” Proceedings of the 12th international workshop on Network and operating systems support for digital audio and video, pp. 53–61, 2002. [7] Z. Wang, H. S.

Savci, and N. S. Dogan, "1-V ultra-low-power CMOS LC VCO for UHF quadrature signal generation," in Circuits and Systems, 2006. ISCAS 2006. Proceedings. 2006 IEEE

International Symposium on, 2006, p. 4 pp.

[20] W. Jiang, J. Lennox, H. Schulzrinne and K. Singh, “Towards Junking the PBX:

Deploying IP Telephony,” NOSSDAV, pp. 177–185, 2001.

[21] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston, J. Peterson, R. Sparks, M.

Handley, and E. Schooler, “SIP: Session Initiation Protocol,” IETF RFC 3261, June 2002.

[22] M. Handley, andV. Jacobson, “SDP: Session Description Protocol,” IETF RFC 2327, April 1998.

[23] A.B. Roach, “Session Initiation Protocol (SIP) Specific Event Notification,”

RFC 3265, June 2002.

[24] SIP for Instant Messaging and Presence Leveraging Extensions (SIMPLE) IETF Working Group. http://www.ietf.org/html. charters/simple-charter.html

[25] B. Campbell, R. Mahy, C. Jennings , “The Message Session Relay Protocol,” Internet Draft (Work in Progress) draft-ietf-simplemessage-sessions-07.txt , July 2004.

[26] M. Colajanni, P.S. Yu and D.M. Dias, “Analysis of Task Assignment Policies in Scalable Distributed Web-Servers Systems,” IEEE Transactions on Parallel and Distributed Systems, Vol. (, N). 6, June 1998.

[27] V. Cardellini, E. Casalicchio, M. Colajanni, and P.S., Yu, “The State of the Art in Locally Distributed Web-server systems,” ACM computing Surveys, vol. 34, no. 2, pp.

263-311, June 2002.

[28] Vivel S. Pai, M. Aron, G. Banga, M. Svendsen, P. Druschel, W. Zwaenepoel, and E.

Nahum, “Locality-Aware Request Distributed in Cluster-base Network Servers,” Proc.

ACM Arch. Support for Progr. Languages Conf., pp. 535-544, October 1998.

[29] E. Casalicchio and M. Colahanni, “A Client-Aware Dispatching Algorithm for Web Clusters Providing Multiple Services,” Proc. ACM World Wide Web Conf., pp.

535-544, May 2001.

[30] SESSION BORDER CONTROLLERS ENABLING THE VOIP

[31] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston, J. Peterson, R. Sparks, M.

Handley, and E. Schooler, “SIP: Session Initiation Protocol,” IETF RFC 3261, June 2002.