تجزیه و تحلیل هزینه و بهینه سازی برای مدیریت گروه چندپخشی IP
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|23362||2007||10 صفحه PDF||سفارش دهید||6580 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Computer Communications, Volume 30, Issue 8, 8 June 2007, Pages 1721–1730
This paper studies the system parameters that affect the total cost of managing the multicast group on a router. A Petri net model is first proposed to describe the states and transitions of the multicast group management. Based on this model, a delayed vacation queue, extended from a simple M/M/1 queue, is used to analyze the total cost of the multicast group management under various system conditions. The formal analysis reveals that the total cost of the multicast group management is minimized when a router delays a certain time to send its pruning messages to upstream routers. Furthermore, a formula is derived to calculate the optimal delay time for sending the pruning messages to minimize the multicast group management cost. To validate this cost analysis model, an IP multicast group management module is added to the NS-2 simulator to simulate the IP multicast group management. The simulation results under various system parameters are consistent with the results obtained by formal analyses, validating the proposed cost analysis model.
Internet Protocol (IP) multicast  and  is a bandwidth-conserving technique that reduces network traffic by simultaneously delivering a single stream of information to a large number of network hosts. IP multicast delivers source traffic to multiple receivers without adding any additional burden on the source or the receivers while using the least network bandwidth of any competing technology. High-bandwidth applications, such as video streaming, may require a large portion of the available network bandwidth for a single stream. In these applications, an efficient way of sending data simultaneously to more than one receiver is by using IP multicast. Fig. 1 demonstrates how data are delivered from one source to a group of interested recipients by using IP multicast. Full-size image (17 K) Fig. 1. IP multicast. Figure options Multicast is based on the concept of grouping the receivers with a common interest to receive a single data stream. This multicast group should not have any physical or geographical boundaries, i.e., the hosts can be located anywhere on the Internet. Although senders not belonging to a multicast group can send datagrams to group members, hosts must be a member of this group to receive the multicast data stream. Since any host can join a multicast group without authentication and leave the group at any time, it is challenging to efficiently manage the dynamic multicast group. In current Internet architecture, the management and maintenance of the multicast group membership is defined in the Internet Group Management Protocol (IGMP) ,  and . IGMP is a protocol used between hosts and multicast routers on a single physical network to establish membership of a particular multicast group. A host identifies its group membership by sending IGMP messages to its associated router and the router, on the other hand, listens to IGMP messages and periodically sends out queries to discover which groups are active or inactive on a particular subnet. Multicast routers use the group information, in conjunction with a routing protocol , ,  and , to support IP multicast over the Internet. Most of recent multicast studies focus on application layer multicast protocols ,  and  or topology aware multicast protocols , , , ,  and  due to the emergence of new applications, such as P2P and wireless ad hoc networks. Although there are deployment issues in IP multicast services and architectures , IP multicast is still the most cost-effective technique in the current Internet architecture to deliver source traffic to multiple receivers because it does not add any additional burden on the source or the receivers . Thus, it is critical to study the cost efficiency of the IP multicast routing and group management, and investigate how to achieve the best IP multicast performance with the least cost in routers. Recently, only a few studies  and  tried to analyze the multicast performance. However, these studies focus on design different methods or tools to measure the performances of multicast protocols. No effort was made to evaluate how IGMP protocols affect the performance of the IP multicast, nor did they try to improve the performance of the IP multicast. Regarding improving the IP multicast performance, Gorinsky et al.  discussed the possible solutions for designing robust multicast congestion control protocols for IP multicast in the presence of untrusted hosts. Nonetheless, none of these existing studies evaluate the impact of IP multicast group management on the performance of the IP multicast. In this paper, we evaluate how the delay of pruning the IP multicast tree impacts the cost of IP multicast group management and determine the optimal delay time for IP multicast group management under various system conditions. We first propose a Petri net to model the IP multicast group management and, in turn, develop a cost analysis model using a delayed vacation queue, extended from a simple M/M/1 queue, to evaluate the total cost of multicast group management on a router . Using this delayed vacation queuing model, we derive a formula to calculate the optimal delay time for a router to send pruning message to its upstream routers so that the cost of multicast group management is minimized. Our simulation studies validate our cost analysis and optimization model. The rest of the paper is organized as follows. After discussing our proposed Petri net model for the multicast group management in Section 2, we analyze the total cost of multicast group management on a router, using a delayed vacation queuing system, under various system conditions in Section 3. In Section 4, we use simulation studies to validate our cost analysis model. Finally, we present our concluding remarks and discuss future work in Section 5.
نتیجه گیری انگلیسی
In this paper, we study the system parameters that affect the total cost of managing the multicast group on a router. We first propose an event/condition Petri net to model the states and transitions of IP multicast group management. Based on this model, we propose a delayed vacation queue, extended from a simple M/M/1 queue, to analyze the total cost of the IP multicast group management under various system conditions. Our formal analysis shows that the total cost of the multicast group management is minimized when a router delays a certain time to send its prune messages to the upstream router. We further derive a formula for calculating the optimal delay time of sending the prune messages. Such a formula can be easily integrated into any IP multicast group management schemes to minimize the total cost of the IP multicast group management. To validate our analysis model, we extend the NS-2 simulator with an IP multicast group management module to evaluate the impact of different system parameters to the total cost of the IP multicast group management. The simulation results show that our cost analysis method is practical. We are currently integrating our proposed cost optimization strategy into the IGMP protocol and study the optimal cost of a router in a real multicast system. In addition, we will study the cost issues in application layer multicast using the similar approach.