A Distributed Scheme for Value-Based Bandwidth Reconfiguration.
ABSTRACT ûAKE: Below I changed and added a few words to include the notation of a "band- with broker". This paper presents a scheme for reallocating bandwidth in path-oriented trans- port networks based on, for example, MPLS, ATM, SDH or SONET. At specified time points, bandwidth can be allocated to those paths that (possibly temporarily) value it most highly. The scheme resembles the bandwidth broker concept in that it deploys trading, but different in that it acts according to local rules and without centralised control. The proposed scheme is thus distributed and scalable. We present a simulation model of a 30-node 46-link network which we use to evaluate the efficacy of the bandwidth reallocation scheme. Th e simulation study shows that bandwidth reconfiguration can substantially reduce both the connection loss probabilities and the amount of lost revenue.
- SourceAvailable from: satnac.org.za[show abstract] [hide abstract]
ABSTRACT: This paper presents a model of network failure recovery based on a scheme for reallocating bandwidth in path-oriented transport networks. We first describe the bandwidth reallocation scheme where, at specified time points, bandwidth can be switched to those routes that (possibly temporarily) value it most highly. This is done entirely according to local rules and without centralised control.We next describe our model of network failure recovery. We consider a protection switching recovery model which works with pre-established reserve-on-demand recovery routes. A reserve-on-demand recovery route allocates required resources after a failure on the working path has been detected. The bandwidth reallocation scheme is used to distribute bandwidth between the working routes before and after a failure, and between the working and recovery routes during the recovery cycle.To support the efficacy of such a system, we present a simulation model of failure recovery in a 30-node 46-link network. The simulation study reveals that bandwidth reallocation allows rapid recovery after the failure of a single link. Traffic is restarted on a recovery route within one round trip time and the recovery path attains the same Grade of Service (GoS) as its working counterpart within 0.1 call holding times.
Conference Proceeding: Implementation issues on market-based QoS control[show abstract] [hide abstract]
ABSTRACT: We discuss two major tradeoffs, spatial and temporal tradeoffs, that appear when applying market-based computing to multimedia network applications. The former appears between computation and communication cost, depending on how agents are distributed over a network. The latter appears between reactiveness and correctness of a result, depending on how the network environment dynamically changes. By implementing a market-based resource allocation mechanism to a desktop conferencing system, we clarified that: as for spatial tradeoff the centralized computation becomes profitable in proportion to the number of clients; and that as for temporal tradeoff the merit to respond quickly to the change of the environment by prematurely terminating the computation supersedes the merit to improve the accuracy of the resource allocation by performing the calculation until the market perfectly clears. It has also been proved that the market-based mechanism can achieve efficient allocation in an actual network environmentMulti Agent Systems, 1998. Proceedings. International Conference on; 08/1998
- [show abstract] [hide abstract]
ABSTRACT: A large number of network applications require a particular Quality of Service (QoS), that can be provided through proper network resource allocation. Furthermore, certain applications (multimedia oriented) may require guarantees of resource availability for predictable QoS. This paper introduces a distributed multi-market approach to network resource allocation. In this approach link bandwidth is bought and sold in two types of markets: the reservation market and the spot market.04/2002;
A Distributed Scheme for Value-Based
˚ A. Arvidssona, B.A. Chierab, A.E. Krzesinskic,1,∗
and P.G. Taylord,2
aEricsson AB, Development Unit IP Networks, Box 1505 SE-125 25¨Alvsj¨ o,
bDepartment of Electrical and Electronic Engineering, University of Adelaide,
5005 Adelaide, Australia.
cDepartment of Computer Science, University of Stellenbosch, 7600 Stellenbosch,
dDepartment of Mathematics and Statistics, University of Melbourne, 3010
This paper presents a scheme for reallocating bandwidth in path-oriented trans-
port networks. At specified time points, bandwidth can be allocated to those paths
that (possibly temporarily) value it most highly.
The scheme resembles the bandwidth broker concept in that it deploys trading,
but different in that it acts according to local rules and without centralised control.
The proposed scheme is thus distributed and scalable.
We present a simulation model of a 30-node 46-link network which we use to
evaluate the efficacy of the bandwidth reallocation scheme. The simulation study
shows that bandwidth reconfiguration can substantially reduce the connection loss
Key words: bandwidth prices, bandwidth reconfiguration, distributed control,
network planning and optimisation, scalability.
Email addresses: Ake.Arvidsson@ericsson.com (˚ A. Arvidsson),
email@example.com (B.A. Chiera), firstname.lastname@example.org (A.E.
Krzesinski), email@example.com (P.G. Taylor).
1Supported by grants 2054027 and 2677 from the South African National Research
Foundation, Siemens Telecommunications and Telkom SA Limited.
2Supported by the Australian Research Council Centre of Excellence for the Math-
In this paper we discuss a distributed scheme for bandwidth reallocation that
can be used in the context of any path-oriented network in which relatively
long-lived paths are used to provision resources for connections or flows, whose
average holding times are much less than the path lifetimes. Possible envi-
ronments in which such a model could be useful are Multi-Protocol Label
Switching (MPLS) networks in which Label Switched Paths (LSPs) act as the
long-lived paths, or Asynchronous Transfer Mode (ATM) networks in which
Virtual Paths (VPs) act as the long-lived paths. Other examples include the
Synchronous Digital Hierarchy (SDH) or the Synchronous Optical Network
We shall use terminology that does not suggest any of the specific envi-
ronments mentioned above. Thus, we call a long-lived path a route. Routes
traverse one or more physical links. In the present paper we restrict our-
selves to connection-oriented traffic with admission controls. Hence we use the
terms connections and lost connections. Note that connections may be circuit
switched (like traditional voice) or packet switched (like voice over IP). The
number of connections that can be simultaneously carried on a route depends
on the amount of bandwidth allocated to the route. However, at any point
in time, it is possible that the connections in service on one route are using
only a small proportion of the bandwidth allocated to that route, while the
bandwidth on another route is heavily utilised. In such a case, it makes sense
to transfer bandwidth from the first route to the second one, if that is possible.
We will employ a scheme in which each route places a value on bandwidth,
dependent on its current bandwidth and its current occupancy. Bandwidth
can then be transferred from routes that place a low value on bandwidth to
routes that place a high value on bandwidth. Under our proposed scheme,
an automated bandwidth manager is assigned to each route. These managers
essentially act as intra-domain bandwidth brokers, cf. , which operate ac-
cording to simple deterministic rules. The managers use knowledge of the
routes’ current occupancy to perform admission control and to calculate the
value of an extra unit of bandwidth (the “buying price”) and also the value
that the route would lose should it give up a unit of bandwidth (the “selling
price”). The managers then use these prices to determine whether the route
should acquire or release a unit of bandwidth or do neither.
We shall view routes as being of two types: direct routes, which traverse just a
single physical link, and transit routes, which traverse more than one physical
link. We assume that each link supports a direct route. The direct routes on
the links of a transit route are referred to as its constituent direct routes.
ematics and Statistics of Complex Systems.
Bandwidth reallocation is driven by the managers of transit routes. The defin-
ing feature of the scheme is that bandwidth reallocation takes place between
transit routes and their constituent direct routes. In this way the managers
are autonomous, act without centralized control from a system coordinator
and behave entirely according to local rules.
Buying and selling prices are communicated via a signalling mechanism. Sig-
nals or control packets are sent at certain intervals along each transit route,
recording the buying and selling prices of the constituent direct routes. If the
sum of the direct route buying prices is greater than the transit route selling
price, then the transit route gives up a unit of bandwidth, which is taken up
by each of the direct routes. Alternatively, if the sum of the direct route selling
prices is less than the transit route buying price, then each of the direct routes
gives up a unit of bandwidth, which is taken up by the transit route.
The managers of a transit route and its constituent direct routes are aware
of local resource demands and bandwidth prices. These managers reallocate
bandwidth among their respective routes in order to maintain the performance
of their routes. The scheme is thus distributed and scalable.
The purpose of this paper is to demonstrate that the reallocation scheme as
described above can improve the performance of a network. For example, we
will study whether it can increase the average rate of earning revenue or re-
duce the traffic-weighted average blocking probability. In Section 2 we briefly
discuss a method for calculating buying and selling prices for bandwidth. The
bandwidth reallocation scheme is presented in Section 3. Results from numer-
ical experiments to test the efficacy of the reallocation scheme are given in
Section 4, and our conclusions are presented in Section 5. Details concerning
the simulation model and its parameterization are given in Appendix A and
Appendix B respectively.
2The price of bandwidth
In Section 1 we described a situation where a manager would wish to place a
value on the amount of bandwidth that is allocated to a route. The question
arises as to how this value should be calculated.
Much work has been done in the area of bandwidth valuation. For example
Lanning et al.  studied the prices that should be charged to customers in a
dynamic loss system: their perspective was that of an Internet billing system
where arrival rates were user-cost dependent. Fulp and Reeves  concentrated
on multi-market scenarios where the price of bandwidth was determined on
the basis of current and future usage. In the context of rate control, Kelly et
al.  and Low and Lapsley , proposed distributed models that optimise
different types of aggregate utility as seen by sources. In these models, the
price of bandwidth on a particular link was identified with the value of dual
variables in a Lagrangian formulation.
Other models include WALRAS  which computes prices for bandwidth
trading in a market-oriented environment by use of a tatonnement process
in a competitive equilibrium. This model is set up as a producer-consumer
system which requires the simultaneous solution of three constrained linear
programming (LP) problems. WALRAS was used  in the design of a system
where bandwidth is traded at prices computed to reflect current and future
requirements. The main disadvantage of the WALRAS model is that the time
for computation can exceed the time in which the underlying market changes.
This problem was overcome, to some degree,  by interrupting the WALRAS
calculation after some adequate timespan and using the interrupted prices.
The model can then be restarted with the updated network information only
to be interrupted again at some further point in time.
An alternative approach to the pricing of bandwidth is presented in . For
an Erlang loss system, this method computed the expected lost revenue due
to connections being blocked, conditional on the system starting in a given
state. The expected lost revenue was used to compute both a buying price and
selling price for bandwidth, relying on knowledge of the state at time zero. We
shall use the bandwidth value calculation presented in , which is described
in more detail below.
Let us consider a single service, such as voice, the holding time of which
amounts to one bandwidth unit and the bandwidth consumption of which
amounts to one capacity unit per connection. Furthermore, assume that a
route manager is making decisions for a planning horizon of τ time units.
Then we can regard the value of U extra units of bandwidth, or the buying
price, as the difference in the total expected lost revenue over time [0,τ] if the
route were to increase its bandwidth by U units at time zero. Conversely, we
can calculate the selling price of U units of bandwidth as the difference in the
total expected lost revenue over time [0,τ] if the route were to decrease its
bandwidth by U units.
For a route with capacity C, let Rc,C(τ) denote the expected revenue lost in
the interval [0,τ], given that there are c connections in progress on the route
at time 0. Then the buying and selling prices Bc,C(τ,U) and Sc,C(τ,U) of U
units of bandwidth when the initial state is c, the current bandwidth is C and
the planning horizon is τ are given by
Bc,C(τ,U) = Rc,C(τ) − Rc,C+U(τ)
?Rc,C−U(τ) − Rc,C(τ),
RC−U,C−U(τ) − Rc,C(τ),
c ≤ C − U
C − U < c ≤ C.
The second equation in (1) was not explicitly given in  for the case C −U <
c ≤ C. In this case, if the route were to give up U units of bandwidth, it would
also have to eject one or more of its current customers. The issue then arises as
to whether an extra “penalty” value should be added to reflect the negative
consequences of such a decision. The right hand side of eq. (1) reflects the
situation in which no such penalty is added. The opposite extreme would be to
incorporate an “infinite” penalty, which would have the effect of precluding any
bandwidth reallocation that requires the ejection of connections. For the rest
of this paper, we adopt the latter course. That is, we do not allow bandwidth
reallocation away from a route with more than C − U customers.
In  it was shown that Rc,C(τ) is given by the inverse of the Laplace transform
where λ and µ are the parameters of the exponential connection arrival and
connection service processes respectively, θ is the expected revenue earned per
(s + Cµ)PC(s/λ) − CµPC−1(s/λ)
Pc(s/λ) = (−µ/λ)cΓ(λ/µ)
is a Charlier polynomial as defined in . For all c and C, the function Rc,C(τ)
is a concave function of τ . It is defined only for integer values of C, but,
for all c and τ, is a strictly convex function of C in the sense that, for all U,
Bc,C(τ,U) < Sc,C(τ,U). (2)
In  a computationally stable method for efficiently calculating the func-
tions?Rc,C(s) was developed. These functions can be inverted using the Euler
oped in , can be used.
method . Alternatively, a linear approximation of Rc,C(τ) itself, also devel-