R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," in Proc. IEEE Infocom, 2000, pp. 1166--1175.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... is made such that the sum of (weighted) proportional gains cannot be increased [10, 11] This notion has generated momentum and work has been done to show that proportional fairness can be achieved within current protocols with the use of both fixed window and dynamic window adjustment schemes [13, 17]. In computational resources, this notion of proportional allocation also exists as the lottery scheduling mechanisms partition resources in direct proportion to the tickets owned, where the tickets are analogous to weights. Because of the ease of implementability, proportional share systems are ....

La, R. J. and V. Anantharam: 2000, `Charge-sensitive TCP and rate control in the Internet'. In: Proceedings of IEEE INFOCOM. pp. 1166--1175.

....of resources for multirate multicast sessions. However, most of the work in this area is concerned only with the notion of max min fairness (see [23] 25] 26] 27] 9] Although there has been a lot of research on the utility maximization problem for unicast case [14] 2] 17] 28] [16], 10] the multirate multicast case has not received significant attention. It is worth noting here that certain factors make the multirate multicast problem significantly different and considerably more complex than its unicast version. For instance, the problem in the multirate multicast case ....

....An aggregate utility maximization approach to flow control was suggested recently by Kelly [13] Recently, this problem has received considerable attention in the context of unicast networks. Several flow control algorithms, both rate based and window based, have been proposed (see [17] 14] [16], 28] 10] These algorithms were derived using different optimization approaches, and we will not discuss them here. Amongst these, the unicast algorithm presented in [10] is also based on subgradient optimization methods. For the special case of all unicast sessions, the algorithm presented ....

R. La, V. Anantharam,"Charge-Sensitive TCP and Rate Control in the Internet ", Proceedings of Infocom 2000.

....peak traffic time periods and low rates when large radio resources are available. Thus, making prices dependent on the network usage can be an efficient solution to network congestion problems. Dynamic pricing has been mainly used to control wired networks supporting Internet based services [5] [6], 7] In this case, techniques to derive the system optimal rate have been proposed, which charge users on the basis of the congestion they cause to the network. In [8] dynamic pricing has been applied to the wireless environment and simulation results have been presented for a simple pricing ....

R.J. La, V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," IEEE INFOCOM 2000.

....not due to congestion may lead to the starvation of TCP Vegas. The authors in [16] conjecture that TCP Vegas may benefit from the use of AQM schemes and present simulations showing the point. The conjecture is based on a differential equation model. Charge sensitive TCP, with milestone works as [17, 18], is also often cited as a way to improve TCP performance. These schemes aims at finding a stable operation point for the network, disregarding the transient behavior (similarly to the work presented in [13] during which the schemes cannot guarantee a low loss rate; however, Internet traffic is ....

R.J. La, V. Ananthram. Charge-Sensitive TCP and Rate Control in the Internet. In IEEE INFOCOM 2000.

....solve the optimization problem, however, it is not a requirement and thus the charge can be viewed as virtual . the most resources in the net will get the lowest rate. By choosing the willingness to pay ws one can give weights to each connection. The rates are then weighted proportionally fair [9] or equivalently the rates per unit charge are proportionally fair [6] x; x8 0 (7) 28 3 Implementation 3.1 Source Algorithm The purpose of the source algorithm is to adjust the rate based on pricing information and the user s willingness to pay. The implicit assumption of the theory that ....

R. La and V. Anantharam. Charge-Sensitive TCP and Rate Control in the Internet. Proceedings of the IEEE Infocorn '00, March 2000.

....the usage of the network, we can make a better use of the available bandwidth, and provide the desired quality of service to the users as well as a greater revenue to the network operator. Dynamic pricing has been mainly introduced to control wired networks supporting Internet based services [5] [6], 7] In this case, techniques to derive the system optimal rate have been proposed, which charge users on the basis of the congestion they cause to the network. In [8] dynamic pricing has been applied to the wireless environment and simulation results have been presented for a simple pricing ....

....the optimal one. Further investigation is needed to solve the pricing problem to optimum. IV. FUTURE WORK ON PRICING FOR DATA PACKET SERVICES In this section, we consider the possibility to apply pricing strategies to connection less services. This problem was already addressed in [5] [6], 7] 10] 11] We consider different classes of users according to their de 1e 05 0.0001 0.001 0.01 0.1 1 P b (Blocking Probability) r (Traffic Load) Linear vs. Flat rate Pricing (N = 8) Fig. 3. Linear vs. Flat rate: call blocking probability vs. traffic load for different ....

R.J. La, V. Anantharam, "Charge-sensitive TCP and rate control in the Internet, " IEEE INFOCOM 2000.

....routes. Considering this scenario, we pose the following problem: given a required network lifetime, what is the most beneficial source rate allocation and flow control strategy To answer this question, we draw upon previous work on congestion pricing in wired networks [8] 9] 10] [11], 12] 13] 14] Their approach consists in deriving the control schemes for the sources traffic rate as solutions of an optimization problem. Each traffic source is associated with a utility function increasing in its transmission rate and subject to bandwidth constraints; the network ....

....are higher than that obtained through the ORSA scheme. This shows that the ORSA algorithm balances the load in the network, which results in a more equitable use of energy resources. V. RELATED WORK Optimization based flow control schemes have been previously proposed in [8] 9] 10] [11], 12] 13] 14] 15] 16] in the context of wired networks. Most of the work on flow control has been done under the assumption that a single path exists between each source destination pair. In [9] 10] 13] each traffic source is associated with a utility function increasing in its ....

R. La, V. Ananthram, "Charge Sensitive TCP and Rate Control in the Internet," Proc. of IEEE INFOCOM 2000.

....they currently receive. Convergence to Nash equilibria [1] is studied under the assumption of synchronous user updates of their access control parameters. The results obtained herein can be extended to asynchronous ( decentralized ) updates using techniques and assumptions described in [15]. A special case of our generic user access mechanism is TCP s slow start congestion control policy (additive rate increase and multiplicative decrease) extended in a natural way to include differential and targeted user demands [13] Our previous work on ALOHA [9] is another application of this ....

.... Consider the following modification of the game (1) without the constraints # n : #(F (# ) 8) for a fixed small # 0 where the n component of the function F : 0, N# [0, is F n (#) G(y n ,# n (#) # n ) This is simply the Jacobi update scheme, see equation (15) of [15]. For small #, we can approximate (8) by the continuoustime game #(t) F (#(t) #(t) 9) The Jacobi iteration will be used to show the convergence of the fixed point iteration. In general, it is hard to show the fixed point iteration converges to the fixed point due to the nonlinearity of # ....

R. La and V. Anantharam. Charge sensitive TCP and rate control in the internet. In Proc. IEEE INFOCOM, 2000.

....[3] and noncooperative [4] 5] 6] frameworks. Congestion control schemes utilizing pricing schemes based on explicit feedback have been proposed by Kelly et al. 7] 8] and Gibbens et al. 9] and subsequent studies have further elaborated on this approach following its basic principles [10] [11], 12] Research supported in part by the NSF grant CCR 00 85917. All correspondence should be forwarded to: Prof. Tamer Basar, Coordinated Science Laboratory, University of Illinois, 1308 West Main Street, Urbana, IL 61801 2307 USA. Tel: 217) 333 3607; Fax: 217) 244 1653; E mail: ....

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in INFOCOM (3), 2000.

....in turn based on the sending rate information fed back from all the sources using this link. Since only a relaxation of the original optimization problem is solved, it is not proven that the algorithm converges to the optimal tra#c allocation. The same formulation is used in the work by La, et al. [9] where the algorithm provided is proven to converge to the optimum rate allocation in the single congested link case (low tra#c networks) Low [10] uses a technique which converts a constrained problem into a non constrained dual problem. This reformulation results in a similar distributed ....

....if one would also have MRGS, UBRS and MRGUBS CoS present. In the simulations described below, we use link full as congestion indicator for all CoSs. So a single binary feedback is used for all CoSs. 5. 1 The Network Topology: The model of the network for our simulations is based on the one in [9] and is shown in Figure 2, where we also show all the links bandwidths and delays, as well as source and destination nodes. The key di#erence, however, is that in our setup we allow for multiple paths for each type of calls. Moreover, we also allow for multiple CoSs. We consider a total of n = 8 ....

R. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," in Proceedings of IEEE INFOCOM'2000.

....routes. Considering this scenario, we pose the following problem: given a required network lifetime, what is the most beneficial source rate allocation and flow control strategy To answer this question, we draw upon previous work on congestion pricing in wired networks [8] 9] 10] [11], 12] 13] 14] Their approach consists in deriving the control schemes for the sources traffic rate as solutions of an optimization problem. Each traffic source is associated with a utility function increasing in its transmission rate and subject to bandwidth constraints; the network ....

....are higher than that obtained through the ORSA scheme. This shows that the ORSA algorithm balances the load in the network, which results in a more equitable use of energy resources. V. RELATED WORK Optimization based flow control schemes have been previously proposed in [8] 9] 10] [11], 12] 13] 14] 15] 16] in the context of wired networks. Most of the work on flow control has been done under the assumption that a single path exists between each sourcedestination pair. In [9] 10] 13] each traffic source is associated with a utility function increasing in its ....

R. La and V. Ananthram, "Charge Sensitive TCP and Rate Control in the Internet," Proc. of IEEE INFOCOM 2000.

....population, we show that the resulting allocation is weighted max min fair, where the weights associated with a user is a function of its utility function. Price based resource allocation schemes, and the analysis of their fairness properties, have received much attention in recent years ( 4] [5], 6] and [10] Broadly, this research can be classified into two categories. a) Market Based Approach: In this approach, utility functions are used to characterize the service requirements of an individual user. It is assumed that the utility functions of individual users are unknown to the ....

....rate to optimize their net benefit. In [4] it is shows that the resulting scheme maximizes (in equilibrium) the social welfare and achieves a (weighted) proportional fair allocation of the network resources. Using similar ideas as presented in the work in [4] La and Anatharam extend in [5] the work of Mo and Walrand [10] and propose a family of window based allocation schemes. They show that their scheme can be used to achieve a (weighted) proportional fair allocation or approximate arbitrarily close a max min fair allocation. b) Control Based Approach: In this approach, the ....

R. J La and V. Anatharam, "Charge-Sensitive TCP and Rate Control in the Internet," prepring, 2001.

....equilibrium (Nash Equilibrium) results in fair resource allocations. The use of pricing to carry out trac management and congestion control (of all sorts [24] has been discussed in the seminal works of Kelly [25] There are other attempts to extend Kelly s ideas in di erent ways [27, 49]. However, the pricing, especially of Internet bandwidth, is usually driven by business reasons rather than academic reasons. Users as well as ISPs often prefer simple at pricing over other complex pricing mechanisms [46] Game Theory has been applied to communication networks in the domain of ....

R. La and V. Anantharam. Charge-sensitive TCP and rate control in the Internet. In Proceedings of INFOCOM, Tel Aviv, Israel, Mar. 2000.

....fair allocation of resources for multirate multicast sessions. However, most of the work in this area is concerned only with the notion of max min fairness (see [21] 23] 24] 25] Although there has been a lot of research on the utility maximization problem for unicast case [12] 15] 26] [14], 9] the multirate multicast case has not received significant attention. It is worth noting here that there are a number of factors which make the multirate multicast problem ISR TECHNICAL REPORT 2000 52 3 significantly different and considerably more complex than its unicast version. 1 For ....

....utility maximization approach to flow control was suggested recently by Kelly [11] This problem, for the case of all unicast sessions, has received considerable attention in recent literature. Several flow control algorithms, both rate based and window based, have been proposed (see [15] 12] [14], 26] 9] These algorithms were derived using different optimization approaches, and we will not discuss them here. Amongst these, the unicast algorithm presented in [9] is also based on subgradient optimization methods. For the special case of all unicast sessions, the algorithm presented in ....

R. La, V. Anantharam,"Charge-Sensitive TCP and Rate Control in the Internet", Proceedings of Infocom

....designed primarily to reduce loss (not increase throughput) and hence, provide balance against the effects of the recovery mechanism. 7 Related Work The optimal flow control problem (4 5) is formulated in [18] It is solved using a penalty function approach in [20, 21] and extended in [30] and [22]. The problem is solved using a duality approach in [25] leading to an abstract algorithm whose convergence has been proved in asynchronous environment. A practical implementation of this algorithm is proposed in [4] The duality approach is extended to multirate multicast setting in [17] The ....

Richard La and Venket Anantharam. Charge-sensitive TCP and rate control in the Internet. In Proceedings of IEEE Infocom, March

.... one byte in the data packet ACK packet header should be sufficient to carry the network congestion feedback (note that one byte would allow 255 links on a user s path) Several alternative approaches to this system utility maximization problem have been proposed in recent literature [9] 8] 10] [11]. These approaches are surveyed in Section VII, where we also compare them with our approach. It is also worth noting here that whereas all these existing approaches require strict concavity of utility functions for guaranteed convergence, we require only concavity. Thus linear utility functions ....

....but different, approach is proposed in [10] In this approach, the user adjusts its rate based on the proportion of marked packets or end to end (measurable) losses. However, the convergence has been proved under certain simplifying assumptions, some of which are not likely to hold in practice. In [11], the authors present a window based flow control approach for the same problem. Here the users choose some weights and the window based flow control scheme, on convergence, allocates rates that are proportionally fair with respect to those weights. The user chooses the next set of weights based ....

[Article contains additional citation context not shown here]

R. La, V. Anantharam,"Charge-Sensitive TCP and Rate Control in the Internet ", Proceedings of Infocom 2000, March 2000.

....an unicast source updates its rate based on the congestion information ( congestion price ) communicated to it by the network. The network updates the congestion information based on the source rates, either communicated by the source or measured. For some other approaches in the unicast case, see [9], 18] As already mentioned, in this paper we consider the problem of maximizing the aggregate receiver utility for the case of multirate multicast sessions. The solutions that we propose are distributed and scalable, and are practical for implementation in a large network. Like the approach ....

R. La, V. Anantharam,"Charge-Sensitive TCP and Rate Control in the Internet ", Proceedings of Infocom 2000, March 2000.

....an unicast source updates its rate based on the congestion information ( congestion price ) communicated to it by the network. The network updates the congestion information based on the source rates, either communicated by the source or measured. For some other approaches in the unicast case, see [9], 18] As already mentioned, in this paper we consider the problem of maximizing the aggregate receiver utility for the case of multirate multicast sessions. The solutions that we propose are distributed and scalable, and are practical for implementation in a large network. Like the approach ....

R. La, V. Anantharam,"Charge-Sensitive TCP and Rate Control in the Internet", Proceedings of Infocom 2000, March 2000.

....of congestion pricing Ayalvadi Ganesh Koenraad Laevens Microsoft Research, 1 Guildhall Street Cambridge CB2 3NH, UK. Richard Steinberg Judge Institute of Management Studies University of Cambridge Cambridge CB2 1AG, UK. June 2000 Technical Report MSR TR 2000 70 Microsoft Research Microsoft Corporation One Microsoft Way Redmond, WA 98052 Abstract We consider the use of congestion pricing as a means for achieving a fair and ecient sharing of bandwidth among end users in communication networks. We propose a model ....

....of congestion pricing Ayalvadi Ganesh Koenraad Laevens Microsoft Research, 1 Guildhall Street Cambridge CB2 3NH, UK. Richard Steinberg Judge Institute of Management Studies University of Cambridge Cambridge CB2 1AG, UK. June 2000 Technical Report MSR TR 2000 70 Microsoft Research Microsoft Corporation One Microsoft Way Redmond, WA 98052 Abstract We consider the use of congestion pricing as a means for achieving a fair and ecient sharing of bandwidth among end users in communication networks. We propose a model for rate adaptation by end ....

[Article contains additional citation context not shown here]

La, R.J. and V. Anantharam, (2000): \Charge-sensitive TCP and rate control in the Internet", Proceedings of Infocom 2000, to appear.

No context found.

Richard La and Venkat Anantharam. Charge-Sensitive TCP and Rate Control in the Internet. In Proceedings of the IEEE Infocom 2000.

No context found.

R. J. La and V. Ananthram, "Charge-sensitive TCP and rate control in the internet," in Proc. IEEE INFOCOM, vol. 3, Mar. 2000, pp. 1166--1175.

No context found.

R. J. La and V. Anantharam. Charge-Sensitive TCP and Rate Control in the Internet. In Proc. IEEE INFOCOM '00, March 2000.

....window sizes of the users and their rates. They have proposed ( proportional fairness, which generalizes maxmin and proportional fairness. Based on this new definition of fairness, they have designed a family of algorithms that achieve ( proportional fairness. La and Anantharam [13] have considered the case where users adopt Mo and Walrand s ( 1) proportionally fair algorithm and are allowed to update their parameters , based on their utilities. They have shown that the system optimum is achieved at the unique equilibrium, which is a fixed point of the mapping ....

....the network. All that is required is a modification of the already existing TCP at the end hosts. We demonstrate, using a fluid model, that at an equilibrium of the algorithm the resulting rates are the optimal rates that solve (# . B. Algorithm The algorithm we have proposed in [13] requires that the users # explicitly compute the optimal prices based on the current prices per unit flow and use the ( algorithm to solve the NETWORK(A, C; p) problem with the given prices . In other words, the problem is formulated as a discrete model where, given the prices per unit ....

[Article contains additional citation context not shown here]

R. J. La and V. Ananthram. Charge-Sensitive TCP and Rate Control in the Internet. In Proc. IEEE INFOCOM '00, March 2000.

....based algorithms for the individual users which converge to window sizes resulting in rates that achieve (p; proportional fairness. Our goal is to find window based algorithms that will solve the system problem of maximizing the aggregate utility of the users. This is as in our earlier work [13] where we had the users adopt Mo and Walrand s (p, 1) proportionally fair algorithm while updating their target queue sizes based on their utilities. In the earlier work, we showed that such an algorithm does achieve the system optimum rates, since it has a unique equilibrium, which is the unique ....

....the network. All that is required is a modification of the already existing TCP at the end hosts. We demonstrate, using a fluid model, that at an equilibrium of the algorithm the resulting rates are the optimal rates that solve SY STEM(U; A; C) B. Algorithm The algorithm we have proposed in [13] requires that the users 16 explicitly compute the optimal prices based on the current prices per unit flow and use the (p, 1) proportionally fair algorithm to solve the NETWORK(A, C; p) problem with the given prices p = p i ; i 2 I) In other words, the problem is formulated as a discrete ....

[Article contains additional citation context not shown here]

R. J. La and V. Anantharam. Charge-Sensitive TCP and Rate Control in the Internet. In Proc. IEEE INFOCOM '00, March 2000.

....update scheme. Theorem 3: In a single bottleneck case, under assumptions (A1) and (A2) the user prices p(n) p 1 (n) p I (n) converge to p = p 1 ; p I ) as n 1 under the totally asynchronous update scheme. The proof of Theorem 3 is omitted in this paper. Refer to [9] for the proof. Let us give a few examples of user utility functions that satisfy the assumptions (A1) A2) Suppose that user i s utility function U i (x i ) has a form 1. U i (x i ) c Delta x a i , where c 0 and 0 a 1 2. U i (x i ) c Delta log(x i b) where c 0 and 0 b 1 ....

Richard J. La and Venkat Anantharam. Charge-Sensitive TCP and Rate Control in the Internet. Full version in preparation.

....occupies at most a small fraction of a resource in the network or when a user cannot correctly compute the price per unit flow as a function of its own p i . In some simple cases, however, users may be able to correctly estimate the effect of their own actions on the prices. This is discussed in [12]. Let us define a mapping T : P P to be T i (p) arg max p i U i ( p i i (p) Gamma p i (14) where P = I , and (p) is the price per unit flow vector at the unique stable point of TCP with price vector p. A fixed point p of the mapping T is a vector in P such that T (p ) ....

R. J. La and V. Ananthram. Charge-Sensitive TCP and Rate Control in the Internet. In Proc. IEEE INFOCOM '00, March 2000.

....Theorem 3: In a single bottleneck case, under assumptions (A1) and (A2) the user prices 6 7 8 :9 converge to ; 8 9 as = 4 under the totally asynchronous update scheme. The proof of Theorem 3 is omitted in this paper. Refer to [9] for the proof. Let us give a few examples of user utility functions that satisfy the assumptions (A1) A2) Suppose that user s utility function has a form 1. A B :C , where AED 0 and 0 FHG F 1 2. A B I HJ , where AKD 0 and 0 . ....

Richard J. La and Venkat Anantharam. Charge-Sensitive TCP and Rate Control in the Internet. Full version in preparation.

No context found.

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," in Proc. IEEE Infocom, 2000, pp. 1166--1175.

No context found.

R J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in INFOCOM (3), 2000, pp. 1166--1175.

No context found.

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," in Proc. IEEE Infocom, 2000.

No context found.

R J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in INFOCOM (3), 2000, pp. 1166--1175.

No context found.

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," in Proc. IEEE Infocom, 2000, pp. 1166--1175.

No context found.

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in Proc. IEEE Infocom, 2000.

No context found.

R J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in INFOCOM (3), 2000, pp. 1166--1175.

No context found.

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in Proc. IEEE Infocom, 2000.

No context found.

R. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," Proc. IEEE Infocom, 2000.

No context found.

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," in Proc. IEEE INFOCOM'2000.

No context found.

Richard La and Venket Anantharam. Charge-sensitive TCP and rate control in the Internet. In Proceedings of IEEE Infocom, March 2000.

No context found.

Richard La and Venket Anantharam. Charge-sensitive TCP and rate control in the Internet. In Proceedings of IEEE Infocom, March 2000.

No context found.

R. La and V. Anantharam, "Charge-sensitive TCP and rate control in the Internet," Proc. IEEE INFOCOM, vol. 3, pp. 1166--1175, 2000.

No context found.

R. J. La and V. Anantharam. Charge-sensitive TCP and rate control in the Internet. In Proceedings of IEEE INFOCOM, pages 1166--1175, 2000. (Cited on pages 35 and 47.)

No context found.

R. La and V. Anantharam. Charge-sensitive TCP and rate control on the internet. In Proceedings of INFOCOM, Tel Aviv, Israel, March 2000.

No context found.

R.J. La and V. Anantharam, \Charge-sensitive TCP and rate control in the Internet," Preprint, 2000, available at http://divine.eecs.berkeley.edu/~ananth/.

No context found.

R. La and V. Anantharam. Charge-sensitive TCP and rate control in the Internet. In Proceedings of IEEE Infocom, March 2000.

No context found.

R. J La and V. Anatharam, "Charge-Sensitive TCP and Rate Control in the Internet," prepring, 2001.

No context found.

R J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in INFOCOM (3), 2000, pp. 1166--1175.

No context found.

R. J. La and V. Anantharam, "Charge-sensitive TCP and rate control in the internet," in Proc. IEEE Infocom, 2000.

No context found.

R. J. La and V. Anantharam, "Charge-Sensitive TCP and Rate Control in the Internet," Proceedings of INFOCOM '2000.

No context found.

R. La, V. Anantharam,"Charge-Sensitive TCP and Rate Control in the Internet", Proceedings of Infocom

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