K. Bharat-Kumar and J. Jaffe, "A new Approach to Performance-Oriented Flow Control", IEEE Transactions on Communications, 29(4):427--435, 1981.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....not just the ones of a single group. 4 While there are other criteria to measure satisfaction such as delay or jitter, bandwidth is a measure of interest to the largest number of applications. 6 Fairness For inter receiver fairness, several measures exist, including the product measure [2] and the fairness index [16] For a discussion of the different measures see [13] Jiang et al. 17] defined inter receiver fairness for a single multicast flow as the sum of the receiver s utilities, where utility is highest around the fair share. Due to the intricacies coming with the utility ....

K. Bharat-Kumar and J. Jaffe, "A new Approach to Performance-Oriented Flow Control", IEEE Transactions on Communications, 29(4):427--435, 1981.

....that weights multicast flows in the allocation with the logarithm of the number of receivers. Second, we measure receiver satisfaction with respect to all receivers, not just the ones of a single group. Fairness For inter receiver fairness, several measures exist, including the product measure [2] and the fairness index [16] For a discussion of the different measures see [13] Jiang et al. 17] defined inter receiver fairness for a single multicast flow as the sum of the receiver s utilities, where utility is highest around the fair share. Due to the intricacies coming with the utility ....

K. Bharat-Kumar and J. Jaffe, "A new Approach to Performance-Oriented Flow Control", IEEE Transactions on Communications, 29(4):427--435, 1981.

....use application B whose performance is much better in the unequal bandwidth allocation. That is, the performance of a collection of applications run by this user over time will be improved. Secondly, pricing may be reduced when bandwidth is reduced. We will return to pricing issues later. In [4], a generalized bandwidth max min flow control algorithm is given for users with different rate characteristics. In the generalization, each flow is associated with a number x p , the throughput priority of user p. At the bottleneck, flow control ensures that a flow s throughput is proportional to ....

K. Bharat-Kumar and J. M. Jaffe. A New Approach to PerformanceOriented Flow Control. IEEE Trans. on Commun., Apr. 1981.

....is (F1) A solution x . 1 , x . 2 , x . N to a flow control problem is fair if for all a A, T(x . a ) 0. Optimality and Fairness 11 This condition is not always satisfied by flow control algorithms that attempt to optimize a global performance metric. For example, in [4] it is shown that maximizing certain global metrics based on power may yield solutions that are not fair by F1. Condition F1 states that no VC should be involuntarily assigned 0 throughput. A VC may decide to terminate if the throughput it is assigned does not meet its requirements. A stronger ....

....Sanders algorithm is decentralized and asynchronous, and maximizes the sum of the individual utilities. The algorithm can assign transmission rate 0 to a circuit and is unfair by condition F1. The algorithm also requires some coordination between choice of utility functions. Kadaba and Jaffe [4] proposed a generalized form of power for for representing throughput delay trade offs. This work focused on maximizing a global objective function based on the power obtained by individual VCs, e.g. the sum of the individual powers. The individual VCs did not choose their own definition of ....

Kadaba Bharath-Kumar and Jeffrey Jaffe. A New Approach to Performance Oriented Flow Control. IEEE Transactions on Communication, COM-29(4), April 1981.

.... bu er over ow has been one of the main causes of process failures in distributed systems (see, e.g. 31] The function of ow control algorithms is to alleviate throughput degradation, unfairness, deadlocks, and failures by preventing such situations from arising (see, e.g. 3, Chapter 6] or [4, 10, 15, 17, 22, 24, 25, 32]) In particular, rate based ow control algorithms adjust transmission rates of di erent sessions in an end to end manner, with the objective to optimize network utilization while still maintaining fairness between di erent sessions (see, e.g. 1, 3, 7, 16, 19, 20, 24] The ratebased approach ....

....complexity which is asymptotically optimal for the class of oblivious algorithms. On the other hand, the centralized algorithm Linear achieves optimal convergence complexity, but does not enjoy locality. We emphasize that our work deviates from previous work on rate based ow control (e.g. [3, 4, 15, 16, 17, 19, 20, 21, 22, 24, 25, 32]) carried out in the data networks community in that it adopts the optimistic approach, put forward in the pioneering work by Afek et al. 1] this approach emerges around the update operation, which allows sessions to intermediately receive rates larger than their nal (max min fair) rates. We ....

K. Bharath-Kumar and J. M. Jae, \A New Approach to Performance-Oriented Flow Control," IEEE Transactions on Communications, Vol. COM-29, No. 4 (Special Issue on Congestion Control in Computer Networks), pp. 427-435, April 1981.

....of the connection with one congested gateway. On the other side, should each connection get the same throughput in pkts sec We consider three different fairness measures: minmax fairness [HG86] the fairness index proposed in [JCH84] and the product measure, a variant of network power [BJ81]. Consider the throughput rate for each connection. To satisfy the min max fairness criteria, the smallest throughput rate must be as large as possible. Given this condition, the next smallest throughput rate must be as large as possible, and so on. For the network in Figure 1, this min max ....

Bharath-Kumar, K., and Jeffrey, J., A New Approach to Performance-Oriented Flow Control, IEEE Transactions on Communications, Vol.COM-29 N.4, April 1981.

....priority share edges, while Linear works with minimum fair share edges. 8 Discussion We have laid out a theory of max min fair, rate based flow control sensitive to priorities of distributed applications, as a significant extension of the classical theory of max min fair, ratebased flow control [2, 3, 18, 20, 21, 22, 31] to the case of networks with guaranteed QoS. Our theory yields an elegant scheme for general prioritized allocation of bandwidth to conflicting distributed applications; this scheme encompasses issues of modeling priorities via priority functions, defining fairness with respect to these ....

K. Bharath-Kumar and J. M. Jaffe, "A New Approach to Performance-Oriented Flow Control, " IEEE Transactions on Communications, Vol. COM-29, No. 4, pp. 427--435, April 1981.

....service (e.g. IP or Unspecified Bit Rate) is that fairness policies can be implemented in resource sharing. A number of policies have been proposed for fairness of a flow control algorithm. For example, one might define fairness weakly to mean that all users get a nonzero share of the resources [1]. Alternatively, one might give preference to traffic that has traveled more hops [6] A commonly used fairness criterion in a rate based flow control system is max min fairness, introduced in [5] This fairness definition has been specified by the ATM Forum as a major goal of an ABR flow control ....

K. Bharat-Kumar and J. M. Jaffe. A New Approach to Performance-Oriented Flow Control. IEEE Trans. on Commun., 1981.

....satisfaction with respect to all receivers, not just the ones of a single group. Last, we use a policy (LogRD) that weights multicast flows in the allocation with the logarithm of the number of receivers. Fairness For inter receiver fairness several measures exist, including product measure [7], and the fairness index[8] for a discussion of the different measures see [9] In [6] inter receiver fairness is defined for a single multicast flow as the sum of the receiver s utilities, where utility is highest around the fair share. Due to the intricacies coming with the utility function we ....

K. Bharat-Kumar and J. Jeffrey, "A new approach to performance-oriented flow control," IEEE Transactions on Communications, vol. 29, no. 4, 1981.

....may not be important during low load when everyone s demands can be satisfied. However, during congestion when the resources are less than the demand, it is important that the available resources be allocated fairly. Defining fairness is not trivial. A number of definitions have been proposed [1,11,15,16]. However, no one definition has been widely accepted. For example, some researchers consider starvation of a few users to be unfair [1] Not allocating any resources to a user is called starvation. By this definition, if all users get a nonzero share of the resources, the scheme is fair. Others ....

....demand, it is important that the available resources be allocated fairly. Defining fairness is not trivial. A number of definitions have been proposed [1,11,15,16] However, no one definition has been widely accepted. For example, some researchers consider starvation of a few users to be unfair [1]. Not allocating any resources to a user is called starvation. By this definition, if all users get a nonzero share of the resources, the scheme is fair. Others argue that a scheme without starvation can still be unfair if the resources are allocated unevenly. The key problem is defining what is ....

K. Bharat-Kumar and J. M. Jaffe, "A New Approach to Performance-Oriented Flow Control, " IEEE Transactions on Communications, Vol. COM-29, No. 4, April 1981, pp. 427-435.

....throughput of the connection with one congested gateway. On the other side, should each connection get the same throughput in pkts sec We consider three different fairness measures: minmax fairness [HG86] the fairness index proposed in [JCH84] and the product measure, a variant of network power [BJ81]. Consider the throughput rate for each connection. To satisfy the min max fairness criteria, the smallest throughput rate must be as large as possible. Given this condition, the next smallest throughput rate must be as large as possible, and so on. For the network in Figure 1, this min max ....

Bharath-Kumar, K., and Jeffrey, J., A New Approach to Performance-Oriented Flow Control, IEEE Transactions on Communications, Vol.COM-29 N.4, April 1981.

....They present a formula defining necessary and sufficient conditions for a set of l i to be Pareto optimal. Finally, they show that in special cases, the Pareto optimal throughputs also maximize the weighted sum of the players powers. The weighted sum of powers has been defined as system power [10, 48] and has been studied in non economic contexts. The flow control economy uses Pareto optimality as the definition of an optimal allocation of resources to the agents in the system. Pareto optimality is a stronger definition for an optimal allocation of resources than the Nash Equilibrium. A Nash ....

....solutions. The simplest condition is (F1) A solution x . 1 , x . 2 , x . N to a flow control problem is fair if for all a A, T(x . a ) 0. This condition is not always satisfied by flow control algorithms that attempt to optimize a global performance metric. For example, in [10] it is shown that maximizing certain global metrics based on power may yield solutions that are not fair by F1. A stronger fairness condition is to allocate the same throughput to all VCs, but differing link capacities can make this impossible. Jaffe [48] proposed a fairness condition that takes ....

[Article contains additional citation context not shown here]

Kadaba Bharath-Kumar and Jeffrey Jaffe. A New Approach to Performance Oriented Flow Control. IEEE Transactions on Communication, COM-29(4), April 1981.

....network utilization and quality of service. Typically, network utilization is quantified in terms of average throughput, and quality of service by average time delay. Maintaining an efficient trade off between throughput and time delay has been traditionally defined as the goal of flow control [KUM81]. Fairness in utilization of network resources has also been proposed as a goal for flow control [GER80, KUM81, BERT92] 1 This confusion is because both flow control and access control are flow regulation controls. The ambiguity about what exactly is meant by the term flow control goes back to ....

....average throughput, and quality of service by average time delay. Maintaining an efficient trade off between throughput and time delay has been traditionally defined as the goal of flow control [KUM81] Fairness in utilization of network resources has also been proposed as a goal for flow control [GER80, KUM81, BERT92]. 1 This confusion is because both flow control and access control are flow regulation controls. The ambiguity about what exactly is meant by the term flow control goes back to the early studies of controls for telecommunication networks: flow control was among the first terms to be ....

[Article contains additional citation context not shown here]

Kabada Bharath-Kumar and Jeffrey M. Jaffe, "A New Approach to Performance Oriented Flow Control," IEEE Transactions on Communications, vol. COM-29, pp. 427--435, April 1981. References 57

....the phase 4 buildup is smaller since the rate probes are more likely to interfere with competing sources than when N = 2. Thus, the peak phase 4 buildup is achieved when N = 2. To summarize, for N in [2,4] there are no losses, and the peak is for N = 2, when there is least interference. For N in [5,7], there are packet losses in phase 2, as well as queue draining in phase 3, and the subsequent phase 4 buildup peaks when N = 7. Finally, the phase 2 buildup peaks when N = 10. N Loss Retransmission Completion Peak Buffer Rate Rate ....

.... 41 12.3. Studies of Ensembles of Controlled Systems One body of work has considered the dynamics of a system where users update their sending rate either synchronously or asynchronously in response to measured round trip delays, or explicit congestion signals, for example in References [5, 7, 8, 13, 52]. These approaches typically assume Poisson sources, availability of global information, a simple flow update rule, and exponential servers. We do not make such assumptions. Further, they deal with the dynamics of the entire system, with the sending rate of all the users explicitly taken into ....

K. Bharath-Kumar and J. M. Jaffe, A New Approach to Performance-Oriented Flow Control, IEEE Trans. on Communication COM-29, 4 (April 1981), 427-435.

....delay and bandwidth cannot be independently allocated. She does not consider GS and BE contracts or the protection property. Douligeris and Majumdar [6 8] have used the concepts of Nash and Stackleberg equilibria to determine optimal flow allocations. However, they try to maximize the total power [4], as opposed to the total utility. They too do not consider pricing. Shenker [24] has used a similar analysis to investigate the properties of Fair Queueing gateways. Economic models of computer systems have been investigated by Yemini [9] and Miller [16] However, they do not consider a game ....

K. Bharath-Kumar and J. M. Jaffe, A New Approach to Performance-Oriented Flow Control, IEEE Trans. on Communication COM-29, 4 (April 1981), 427-435.

....global efficiency. An allocation which is 100 globally efficient and 100 globally fair is said to be globally optimal. It should be pointed out that by associating efficiency with resource power (rather than user power) we have been able to avoid the problems encountered by other researchers [2, 10] in using the power metric. Notice that we have a multi criteria optimization problem since we are trying to maximize efficiency as well as fairness. One way to solve such problems is to combine the multiple criteria into one, for instance by taking a weighted sum or by taking a product. We chose ....

K. Bharat-Kumar and J. M. Jaffe, "A New Approach to Performance- Oriented Flow Control, " IEEE Transactions on Communications, Vol. COM-29, No. 4, April 1981, pp. 427 - 435.

....approaches to flow control have been studied in the past. One body of work has considered the dynamics of a system where users update their sending rate either synchronously or asynchronously in response to measured round trip delays, or explicit congestion signals, for example in references [34 38]. These approaches typically assume Poisson sources, availability of global information, a simple flow update rule, and exponential servers. We do not make such assumptions. Further, they deal with the dynamics of the entire system, with the sending rate of all the users explicitly taken into ....

K. Bharath-Kumar and J. M. Jaffe, A New Approach to Performance-Oriented Flow Control, IEEE Trans. on Communication COM-29, 4 (April 1981), 427-435.

No context found.

K. Bharath-Kumar and J. M. Jaffe, "A New Approach to Performance-Oriented Flow Control," IEEE Trans. Comm., Vol. COM-29, No. 4, April 1981, pp. 427-435.

No context found.

Bharath-Kumar, K., and Jeffrey, J., A New Approach to Performance-Oriented Flow Control, IEEE Transactions on Communications, Vol.COM-29 N.4, April 1981.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC