F. P. Kelly, Charging and Accounting for Bursty Connections, In: Internet Economics, Eds. L. W. McKnight and J. P. Bailey, MIT Press, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... this worst case behavior, a call is admitted if and only if A log m exp s n t n p : 24) The right hand quantity is the ON OFF bound of the effective bandwidth of a source with known mean rate m and peak rate p; the bound is tight when the source is indeed ON OFF (see, e.g. [28]) In general these rates are not known in advance but it is common to provide upper bounds through leaky bucket descriptions which are used to police that source. In the context of ATM the relevant parameters are the sustainable and peak cell rates (SCR, PCR) 29] While the interpretation of ....

F. P. Kelly, "Charging and accounting for bursty connections," in Internet Economics, L. W. McKnight and J. P. Bailey, Eds., pp. 253--278. MIT Press, 1997.

....per packet and per hop accounting) is prohibitive. Indeed, line speeds (the rates at which data packets arrive at and leave from nodes) are generally high relative to the processing power in the switch or router, and so the smart market approach is not implementable with current technology. In [47], users are charged according to a combination of declared and measured characteristics of traffic. By taking an equivalent bandwidth model of resource utilization, and assuming appropriate traffic models, a menu of pricing plans indexed by the declared traffic can be offered which encourages ....

....plans indexed by the declared traffic can be offered which encourages users to make truthful declarations (e.g. of the mean rate) and also encourages the users characterization efforts to be directed where they are most relevant to the network resource allocation. As the pricing is relative, [47] does not aim to address the problem of determining the actual monetary values of the market price (that users would be willing to pay) Thus, the information requirement is only partially reduced, that is, with respect to quantity and not price. Another pricing scheme which incorporates ....

[Article contains additional citation context not shown here]

F. P. Kelly. Charging and accounting for bursty connections. In L. W. McKnight and J. P. Bailey, editors, Internet Economics. MIT Press, 1997.

....Control (CAC, a device that decides whether or not a new request can be admitted to the network, without violating service criteria) This CAC gives rise to the notion of effective bandwidth, an amount of system resources that is occupied by a single connection of a particular type. Kelly [12] proposes to charge a network user based on this effective bandwidth. Organization. This paper is organized as follows. In section 2, we describe both methods that calculate the decay rate of loss probability H n : We emphasize that the first method is valid for both Markov fluid and periodic ....

F. Kelly. Charging and accounting for bursty connections. In: Internet economics, eds. L. McKnight and J. Bailey, 1996.

....which gives the amount of resources that must be reserved for a source in order to satisfy its QoS requirements; such a scalar enables problems (e.g. admission control, routing) in broadband networks carrying bursty traffic to be mapped into problems in traditional circuitswitched networks. In [Kel96b] a unifying definition of the effective bandwidth is introduced. The definition summarizes the statistical characteristics of traffic over different time and space scales, and builds on earlier work on effective bandwidths and asymptotic models. Illustrative examples which demonstrate the ....

....and asymptotic models. Illustrative examples which demonstrate the unifying property of the definition include periodic sources, fractional Brownian input, policed and shaped sources, and deterministic multiplexing. Central to our work is the notion of effective bandwidth, as defined in [Kel96b] and the many sources asymptotics [CW96, BD95] which can compute the space and time parameters that appear in the effective bandwidth definition (see Section 2.1) We apply and evaluate these techniques for performance analysis of broadband links and for quantifying resource usage for real ....

[Article contains additional citation context not shown here]

F.P. Kelly. Charging and accounting for bursty connections. In J. P. Bailey and L. Mcknight, editors, Internet Economics, Massachusetts, 1996. MIT Press.

....[Sairamesh95, Odlyzko99a, Odlyzko99b] and on achieving proportional fairness [Kelly97a,Kelly98] Implementation varies based on the underlying technology. Proposed schemes for ATM networks include use of pricing at connection admission to encourage users to correctly declare QOS parameters [Kelly97b] and use of prices to cause users to exert control over their cell transmission rate [Murphy94] In the Internet, schemes have been proposed that would allow users to place the price that they would be willing to pay for a packet s transmission into the packet s header [MacKie Mason95] The ....

F. Kelly. Internet Economics, chapter `Charging and Accounting for Bursty Connections', pages 253--278. MIT Press, 1997. (p 21)

....to each client. In a shared environment, it is not trivial to ensure that clients resource requirements are always met. Clients perceive performance as low latency, high throughput, and high availability [16] QoS guarantees typically give some assurance of this. We use an existing notion of QoS [13] and improve it to ensure a high probability of resource availability to each client. Related work has been done in the context of stochastic networks [13] Packing and load balancing of individual connection requests (jobs) have been studied on a single network connection. It is based on ....

....latency, high throughput, and high availability [16] QoS guarantees typically give some assurance of this. We use an existing notion of QoS [13] and improve it to ensure a high probability of resource availability to each client. Related work has been done in the context of stochastic networks [13]. Packing and load balancing of individual connection requests (jobs) have been studied on a single network connection. It is based on statistical multiplexing the combining uncorrelated bursty demands to improve packing efficiency. A study of statistical multiplexing from the perspective of ....

[Article contains additional citation context not shown here]

Frank P. Kelly, Charging and Accounting for Bursty Connections, in Internet Economics, Editors Lee W. McKnight and Joseph P. Bailey, MIT Press, 1997, ISBN 0 262 13336 9, pp. 253-278.

....imposed by a packet on other packets ensures that this mechanism provides users with the right incentives to adapt to congestion. The price mechanism studied in this paper differs in an important respect from a number of earlier proposals, which were based on pricing individual flows (see, e.g. [6], 12] It does not require the core network to maintain per flow state. The task of associating prices with end users is pushed to the network edges where it is easier to deal with. We propose a mechanism for users to respond to price information which is motivated by the assumption that they ....

F. P. Kelly, "Charging and accounting for bursty connections", in Internet Economics, Lee W. McKnight and Joseph P. Bailey eds., MIT Press, 1997.

.... of the network [ShCE96] or integrating specialized modules into switches or routers [EdMV95, DeWD97] Alternative approaches define dedicated measurement devices [EdMV95] and statistical sampling methods [Vie90] Different measurement granularity levels can be considered, e.g. virtual connections [Kell96, CoKW97] or TCP flows. Metering of resource consumption can be avoided by limiting usage CHARGING AND ACCOUNTING ARCHITECTURE FOR QOS ENHANCED IP SERVICES 3 metering to metering of resource reservation [FaSP98] or by contractual traffic estimation based on user profiles [ClFa97] Charging ....

F. P. Kelly. Charging and Accounting for Bursty Connections. In "Internet Economics", J. P. Bailey and L. McKnight (editors), MIT Press, Massachusetts, 1996.

....users to choose between different QoS levels corresponding with different prices 3. 2 Effective Bandwidth Approach (Kelly, Courcoubetis et al. This charging method is presented in several papers, each emphasizing a different aspect, by Kelly, Songhurst, Courcoubetis, Siris, Stamoulis, and Weber [8, 9, 10, 11, 12, 21, 22, 23, 45]. Many of the papers are based on research activities performed in the ACTS project CA hMAN. The basic idea is the following. Each connection is charged with a price per until time, a price per bit and an optional connection price . The parameters and depend on the traffic characteristics ....

F. P. Kelly: "Charging and Accounting for Bursty Connections", in Internet Economics, eds. L. W. McKnight, J. P. Bailey, pp 253-278, 1995, http://www.press.umich.edu/jep/ econTOC.html.

....update, but now also projects the new price onto the hyperplane in Theorem 6. We illustrate through an example that this might significantly speed up the convergence towards an equilibrium. 2. 3 Related works Pricing network services and resources have received much attention recently; see e.g. [3, 14, 20, 17, 11, 4, 12][31, Chapter 8] and references therein. This paper differs from the previous work in three respects. First we consider both bandwidth and buffer in our model. Second, and more importantly, we consider both variable and fixed resources where users must balance the higher expected return of variable ....

F. P. Kelly. Charging and accounting for bursty connections. In L. W. McKnight and J. P. Bailey, editors, Internet Economics. MIT Press, 1996.

....in each packet equal to the true valuation. Each node in the network becomes an efficient market, but the engineering cost (sorting packets by bid price, as well as per packet and per hop accounting) could be significant if line speeds are high relative to the processing power in the router. In [14], users are charged according to a combination of declared and measured characteristics of traffic. By taking an equivalent bandwidth model of resource utilization, and assuming appropriate traffic models, a menu of pricing plans indexed by the declared traffic can be offered which encourages ....

....plans indexed by the declared traffic can be offered which encourages users to make truthful declarations (e.g. of the mean rate) and also encourages the users characterization efforts to be directed where they are most relevant to the network resource 3 allocation. As the pricing is relative, [14] does not aim to address the problem of determining the actual monetary values of the market price (that users would be willing to pay) Another pricing scheme which incorporates multiplexing gain is formulated in [12] These and a number of other schemes are summarized in [11] in a comprehensive ....

F. P. Kelly. Charging and accounting for bursty connections. In L. W. McKnight and J. P. Bailey, editors, Internet Economics. MIT Press, 1997.

....as such, but also to compensate for the the blocking nature of the measurement process, which has a negative e ect on throughput. Consequently usage based models are not yet 2 considered viable, and many proposals have focused on compromise solutions based on aggregation [BBC93, Cro96, Cla95, Kel97] As part of a project investigating radical approaches for operational support systems, we are investigating the possibility of lowering the operational cost of usage based charging by shifting responsibility for accounting to the users themselves. We propose that users measure their own trac, ....

Frank P. Kelly. Charging and Accounting for Bursty Connections, pages 253-278. MIT Press, 1997.

....process to the analysis. This allows the methodology to provide the required predictive element. Totest the predictive validity of the m odels extracted, a series of experime nt s w ith real tim e audio and video were conducted. Many networ ks are required to carry bursty traffic [10], whic h can m ean that resulting levels of quality can fluctuate w idely . As ki ng users to m ake QoS evaluations after the interaction precludes the ability to establish w hen during the assessm ent they found the quality unacceptable. Due to these considerations w e developed a sliderinterface ....

Kelly, F. (1997).Charging and accounting for bursty connections. In McKnight, L.W., & Bailey, J.P. (eds.) Internet Economic s. MITPress.

....the network resources, the sel sh users can be steered toward an operating point desired by the network manager, e.g. to better balance the load among communication links. More precisely, pricing can take into account the actually measured usage in terms of trac volume and connection duration [11] [12], 13] or be based on the application s quality of service (QoS) requirement [14] 15] 16] A di erent question is whether the price rates are determined in advance or uctuate depending on the actual state of the network (e.g. current congestion) While the latter can theoretically attain a ....

....to employ a pricing scheme that charges the user for the number of packets sent, as well as connection time. We assume that the connection usage is priced linearly, i.e. it costs a per time unit and b per packet sent; some advantages of this pricing scheme in a di erent context were pointed out in [12], 25] We investigate what the user s strategy of timeouts should be in order to minimize the expected cost of transmitting the information, and independently analyze several models, di ering in whether all the user s packets are routed over the same path or over independent paths in the network. ....

F.P. Kelly. Charging and accounting for bursty connections. In L.W. McKnight and J.P. Bailey, editors, Internet Economics, pages 253-278. MIT Press, Cambridge, MA, 1997.

....of the traffic flow. Brownlee [3] demonstrates feasibility of usage sensitive pricing in a moderately sized experiment. MacKie Mason and Varian [13, 14] argue that usage sensitive pricing could facilitate congestion control as it leads network resources to their most valuable uses. Kelly [11] finds ways to account for burstiness, relying on the concept of effective bandwidths , which can be computed from the traffic characteristics of a variable bit rate stream. The effective bandwidth thus computed lies between the mean and peak rates the higher the burstiness of traffic stream, ....

....the marginal cost. This conclusion also holds with multiple QoS classes. We also compute the costs to individual users. The cost of operating the network must also be recovered, and we are looking for a fair way to assign these costs. This is not a straightforward task; as explained by Kelly [11] the usage of a network resource does not simply coincide with the total number of bits carried. The higher the burstiness of a flow, the more resources should be reserved to handle it. Thus, the allocation of operating costs to users need to reflect the traffic characteristics of the flows ....

[Article contains additional citation context not shown here]

F. Kelly. Charging and accounting for bursty connections. In L. McKnight and J. Bailey, editors, Internet Economics. MIT Press, Cambridge, Massachusetts, 1997.

....with accounting processes as such, but also to compensate for the the blocking nature of the measurement process, which has a negative e ect on throughput. Consequently usage based models are not yet considered viable, and many proposals have focused on compromise solutions based on aggregation [3, 9, 8, 11]. As part of a project investigating radical approaches for operational support systems, we are investigating the possibility of lowering the operational cost of usage based charging by shifting responsibility for billing to the users themselves. We propose that users measure their own trac, and ....

Frank P. Kelly. Charging and Accounting for Bursty Connections, pages 253-278. MIT Press, 1997.

....In comparison with delay systems, there are relatively fewer studies on the pricing of blocking systems. Miller and Buckman [15] focus on transfer pricing of an M=M=N=N system, and conclude it should equal the opportunity cost, which comes from the lost request when all N severs are busy. Kelly [10] considers a network in which the sum of the effective bandwidths of all users can not exceed certain constraints in order to meet statistical performance guarantees, and develops an optimal tariff structure for that network. Wang et al. [16] derive optimal pricing for an integrated services ....

Kelly, F. P. (1997) Charging And Accounting For Bursty Connections, Internet Economics, Lee McKnight and Joseph Bailey eds., The MIT Press, pp. 253-278.

....some of the spare capacity of a VBR channel handling a primary delay sensitive flow. In the VBR channel, the best effort cells could be discerned from the primary cells by flags in the header or payload. Consider an arbitrary connection in a communication network. The user s cost in dollars is [11], 23] ff fiT flN = ff (fi flm)T where ff 0 is the fixed connection set up cost in dollars, T is the total time duration of the call in seconds, N is the total number of cells, and m = N=T is the mean rate of transmission in cells s. The quantities fi and fl will also vary according to ....

F.P. Kelly. Charging and accounting for bursty connections. In Internet Economics (Editors L.W. McKnight and J.P. Bailey), pages 253--278. MIT Press, 1997.

....is crucially needed for commercial service providers, and allows to shape user behavior for achieving common benefits. CA can be based on reserved or on actually used resource or on a combination of both. Usage measurements for charging can be of different granularity (e.g. virtual connections [Kell96, CoKW97], or TCP flows) Cocchi et al. CoES93] introduced a simple priority model with an exemplary price discrimination for different levels of service giving preference to bandwidth and or delay sensitive applications. Their simulation results proof that flat pricing is inferior to a priority model. ....

F. P. Kelly. Charging and Accounting for Bursty Connections, in "Internet Economics", J. P. Bailey and L. McKnight (editors). MIT Press, Massachusetts, 1996.

....by the physical capacity decouples actual use of resources from the monetary charges, making the network vulnerable to the well known tragedy of the commons . Thus there is a need to develop new approaches to pricing of network resources, which has led to much research in recent years (see [11, 8, 7, 3, 12, 9, 2] for a representative sample) 8 th International Symposium on Dynamic Games and Applications, Maastricht, The Netherlands, July 1998. y Corresponding author. In [10] we present the Progressive Second Price (PSP) auction, an efficient mechanism for allocation of variable size shares of a ....

F. P. Kelly. Charging and accounting for bursty connections. In L. W. McKnight and J. P. Bailey, editors, Internet Economics. MIT Press, 1997.

....Varian [18] Usage 8 based pricing is thus a useful tool to induce more efficient sharing of network resources. It may have several components such as an access charge, a connection charge per unit time, a volume charge, and a resource charge. For instance, based on effective bandwidth, Kelly [14] proposes a charging scheme with three components that encourages users to accurately estimate their mean rate, thus allowing more effective sharing of bandwidth. Distributed iterative schemes for setting usage based prices in order to optimize resource allocation have been proposed, e.g. by Low ....

F. P. Kelly, Charging and Accounting for Bursty Connections, In: Internet Economics, Eds. L. W. McKnight and J. P. Bailey, MIT Press, 1996.

....section we present some related work on specific pricing proposals. The nature of this work is primarily not focused towards a globally optimal pricing strategy and focuses rather on efficient pricing strategies for specific traffic profiles or network architectures. 3. 1 Effective Bandwidth [Kelly] presents a method of charging bursty traffic sources based on their effective bandwidth. The effective bandwidth is a function of the traffic profile (or the arrival distribution) of a source and provides a good measure upon which to base pricing decisions. For instance, for an on off source, the ....

....a good measure upon which to base pricing decisions. For instance, for an on off source, the effective bandwidth would be a function of the peak and mean rates. It has been shown that the notion of effective bandwidth is relevant to issues like network dimensioning and call acceptance control. In [Kelly] it has been shown that effective bandwidth is relevant to pricing decisions in a network. The author shows that as long as the effective bandwidth of the source can be expressed as a concave function of the expectation of a measurable quantity, a charging scheme can be developed with certain ....

[Article contains additional citation context not shown here]

Frank P.Kelly; Charging and Accounting for Bursty Connections. in MIT Workshop on Internet Economics, 1995.

....which is a feature of PMP. They also argue for at least some variant of Clark s proposal of charging for expected usage, with those portions of a consumer s offered load that deviate from negotiated statistics being treated at lower priorities. This part is similar to another proposal of Kelly [Kelly]. The problem with charging for negotiated usage profiles, just as with applications of future markets to networks, is that they do not deal with the inevitable short term fluctuations in traffic. It is desirable to provide incentives for users to either lower their load on the network or else ....

F. P. Kelly, Charging and accounting for bursty connections, pp. 253278 in Internet Economics, L. W. McKnight and J. P. Bailey, eds., MIT Press, 1997. Preliminary version in J. Electronic Publishing, special issue on Internet economics, hhttp://www.press.umich.edu/jep/i.

....they receive. Hence, using the testbed, one is able to demonstrate and explain the difference (and, in fact, the advantages) of usage based pricing compared to non usage based pricing. The remainder of the paper is organized as follows: In Section 2, we overview the ABR pricing schemes of [6, 8, 5]. In Section 3, we describe the testbed in detail. In Section 4, we discuss and assess its main features, including the objectives of experiments that can be conducted. Finally in Section 5, we present some concluding remarks. 2 ABR Pricing Schemes The pricing structure for best effort services ....

....for ABR services which have both been developed within the E.C. funded ACTS Project CA hMAN (Charging and Accounting in Multiservice ATM Networks) 5, 4] We briefly describe these schemes hereafter, in order to make the issues to be faced by the testbed more tangible . In the first scheme, [8, 5], prices are based on the Minimum Cell Rate (MCR) and involve measurements of the duration of a connection (Time, T ) and the total number of cells transferred that are above MCR (Volume, V ) The total charge of the connection is a Delta MCR Delta T b Delta V; 1) where a is the charge per ....

[Article contains additional citation context not shown here]

F.P. Kelly. Charging and Accounting for Bursty Connections. In J. P. Bailey and L. Mcknight, editors, Internet Economics, Massachusetts, 1996. MIT Press.

....choice of allocated rates are in equilibrium. 1 Introduction This paper describes a model designed to shed light on the issues of charging, rate control and routing. Its main purpose is to support ongoing work on charging schemes for broadband multiservice networks, described in [3] and [6]. A subsidiary aim is to investigate the relationship between various fairness criteria and smart market approaches to dynamic pricing [7] 9] 10] The organization of the paper is as follows. Section 2 presents a system model of charging, routing and flow control, where the system comprises ....

....described in this paper. A further challenging question concerns how the choice of parameter m s might be implemented in an ATM network. One possibility would be to use the Minimum Cell Rate of ATM standards [5] to buy a share of spare capacity, as well as to provide a lower bound on the rate. In [6] some of the consequential influences on user behaviour are discussed. 2 This deduction is incorrect: it may fail when m s = x s = 0. A counterexample to the Theorem is provided by the example m s = x s = 0, s U 0 s (0) for all s: then statements (i) ii) and (iii) of the Theorem hold, ....

F.P. Kelly, Charging and accounting for bursty connections. In Internet Economics, ed. L.W. McKnight and J.P. Bailey (eds). MIT Press, 1996.

....This work was partly supported by the European Commission under ACTS Project CA hMAN (AC 039) Two charging schemes for ABR services have been developed by the ACTS Project CA hMAN (Charging and Accounting in Multiservice ATM Networks) 4] funded by the European Commission. In the first scheme [10, 4], charges are based on the Minimum Cell Rate (MCR) and involve measurements of the duration of a connection (Time, T ) and the total number of cells transferred (Volume, V ) The total charge of the connection is a Delta MCR Delta T b Delta V , where a is the charge per unit of time and unit ....

F.P. Kelly. Charging and Accounting for Bursty Connections. In J. P. Bailey and L. Mcknight, editors, Internet Economics, pages 253--278, Massachusetts, 1997. MIT Press.

....that the deterioration of QoS (compared to VBR) is accompanied by considerable savings in the charges incurred. Two pricing schemes for ABR services have been developed within the E.U. funded ACTS Project CA hMAN (Charging and Accounting in Multiservice ATM Networks) 4, 3] In the first scheme [10, 4], prices are based on the Minimum Cell Rate (MCR) and involve measurements of the duration This work was partly supported by the European Union under ACTS Project CA hMAN (AC 039) of a connection (Time, T ) and the total number of cells transferred (Volume, V ) the total charge of the ....

F.P. Kelly. Charging and Accounting for Bursty Connections. In J. P. Bailey and L. Mcknight, editors, Internet Economics, Massachusetts, 1996. MIT Press.

.... ae h k e sh k (h k m k (e sh k Gamma 1) 2 oe Gamma m 2 k Gamma e sh k Gamma 1 Delta h k m k Delta C : 21) Information on good choices for m k , k = 1; K may come from a variety of sources, for example users may provide information through their tariff choices [12,17], or the network may have available long term averages for traffic of different types. A.3. Tangent of slope one (A III (n) Suppose that P k m k is known, but not individual values of m k . Can we bound the right hand side of inequality (13) One method leads to the Hoeffding bound [10] ....

F. P. Kelly. Charging and accounting for bursty connections. In L. W. McKnight and J. P. Bailey, editors, Internet Economics. MIT Press, 1996.

No context found.

F. P. Kelly, Charging and Accounting for Bursty Connections, In: Internet Economics, Eds. L. W. McKnight and J. P. Bailey, MIT Press, 1996.

No context found.

. Kelly, F.P , "Charging and Accounting for Bursty Connections," MIT workshop on Internet Economics March 1995.

No context found.

F.P. Kelly, "Charging and accounting for bursty connections". In L.W. McKnight and J.P. Bailey, editors, Internet Economics, pp 253-278, MIT Press, Cambridge MA, 1997.

No context found.

Frank P. Kelly, "Charging and Accounting for Bursty Connections," in: L.W. McKnight and J.P. Bailey, Internet Economics, MIT Press, 1996.

No context found.

Kelly, F. P. (1997). Charging and Accounting for Bursty Connections. In McKnight, L. W. and Bailey, J. P., editors, Internet Economics, pages 253--278. MIT Press. 19 2.12 Traffic Shaping

No context found.

F. Kelly, "Charging and accounting for bursty connections," In L.W. McKnight, and J.P. Bailey, Internet Economics, MIT Press, 1997.

No context found.

F. Kelly, "Charging and Accounting for Bursty Connections", in Internet Economics, J. Bailey and McKnigt eds.

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