IMPLEMENTING A DYNAMIC PRICING SCHEME FOR QOS
ENABLED IPV6 NETWORKS
El-Bahlul Fgee, Shyamala Sivakumar
, W.J. Phillips and W. Robertson, J. Kenny
Department of Engineering Mathematics, Dalhousie University, Halifax, Canada
‡Computing and Information Systems, Sobey School of Business, Saint Mary's University, Halifax
Keywords: IPv6 QoS manager, d
ynamic pricing.
Abstract: Currently the Internet based on IP supports a single best-effort service in which, all packets are queued and
forwarded with the same priority. No guarantees are made regarding timely and guaranteed delivery.
However, many e-commerce applications, that are delay and loss sensitive, use the Internet as a transport
infrastructure because of its reach-ability, and cost efficiency. Challenges faced by ISPs supporting e-
commerce traffic include enhancing their traffic flow handling capabilities, speeding the processing of these
packets at core routers, and incorporating Quality of Service (QoS) methods to differentiate between traffic
flows. These schemes add to infrastructure costs of network providers which can be recovered by
introducing extra charges for QoS enabled traffic. Many pricing schemes have been proposed for QoS-
enabled networks. However, integrated pricing and admission control has not been studied in detail. In this
paper a dynamic pricing model is integrated with an IPv6 QoS manager to study the effects of increasing
traffic flows rates on the increased cost of delivering high priority traffic flows. The pricing agent assigns
prices dynamically calculated according to the network status for each traffic flow accepted by the domain
QoS manager. Combining the pricing strategy with the QoS manager allows only higher priority traffic
packets that are willing to pay more to be processed during congestion. This approach is flexible and
scalable as pricing is decoupled from QoS decisions and reservations.
1 INTRODUCTION
Traditionally IPv4 incorporates simplistic pricing
approaches such as flat rate for unlimited usage and
is based on a fair distribution of costs. IPv6 supports
QoS by introducing the flow label field which
uniquely defines all the packets belonging to the
same flow. In IPv6 domains, a QoS manager can
take advantage of the flow label and traffic class
fields, to handle domain resources and to
differentiate traffic flows. This system is capable of
negotiating QoS parameters, upon accepting user
connection, that includes guaranteeing the quality.
Therefore, the resource for delay and loss sensitive
traffic is secured and these flows are processed faster
than other flows. Such differentiation makes pricing
an important, if not critical issue, in today's Internet.
In the future the issue of pricing will be more
relevant as more E-commerce applications begin to
rely on the Internet for their delivery (Faizuullah,
2000). Flat pricing is unfair as it does not
differentiate between traffic flows with differing
QoS requirements. Proposing appropriate pricing for
QoS-enabled networks is a challenging problem as
pricing must be integrated with admission control
strategies (Tianshu, 2004). Our IPv6 QoS manager
uses a combination of a packet’s flow ID and the
source address (Domain Global Identifier (DGI))
(Fgee, 2004), to process and reserve QoS inside a
domain. This QoS manger is integrated with a
dynamic pricing model (Tianshu, 2003) to study the
effects of increasing traffic rates on the increased
cost of delivering high priority traffic flows. The
pricing agent assigns prices for each traffic flow
initially accepted by the domain manager. Pricing is
dynamic and calculated based on network status.
Section 2 defines QoS and the parameters used to
m
easure it. Section 3 discusses the new elements of
IPv6 needed to enable QoS. Section 4 gives an
289
Fgee E., Sivakumar S., J. Phillips W., Robertson W. and Kenny J. (2005).
IMPLEMENTING A DYNAMIC PRICING SCHEME FOR QOS ENABLED IPV6 NETWORKS.
In Proceedings of the Seventh International Conference on Enterprise Information Systems - DISI, pages 289-292
Copyright
c
SciTePress
overview of the proposed IPv6 QoS manager (Fgee,
2004). Section 5 discusses how the dynamic pricing
model is integrated with the QoS manager.
Simulation results are presented to illustrate the
feasibility of the integrated approach.
2 QOS REQUIRMENTS
QoS is the ability of a network to provide better
services to selected network traffic over different
underlying technologies. End-to-end delay, jitter,
bandwidth and packet loss rate are the parameters
typically used for characterizing QoS of individual
connections or data flows (Jha, 2002). The elements
commonly involved in providing QoS guarantees are
(Jha, 2002): 1) Admission Control determines access
to available network resources and keeps track of all
reservations. 2) Policing is performed when a flow’s
actual data traffic exceeds the requested values given
in the traffic specifications. In such cases the packets
are dropped or downgraded to a best effort service
class or marked as non-conforming. 3) Packet
classification identifies packets belonging to a
specific flow and designates a QoS class for this
flow. 4) Packet scheduler ensures that the flows
identified by the packet classifier receive the
requested QoS. 5) Traffic control implementation
involves queuing methods employed to control
traffic at routers interfaces. Function 1) and 2) are
handled by the domain QoS manager. Functions 3)
and 4) are implemented at the edge points.
3 INTERNET PROTOCOL V6
Two components of IPv6 are used to deliver QoS,
the first is the 8-bit priority field in the IPv6 header
that can be used to identify and discriminate traffic
types based on contents of this field. The second
component is the flow label which is used to label
packets belonging to traffic flows for which the
sender requests special handling. Network elements
can now classify packets based on IP semantics
allowing for efficient mapping of packets to their
flows and hence to their flow specification policy.
The flow label is chosen by the IPv6 QoS manager
and is used for reserving resources, and routing and
monitoring traffic flow packets. Also, packets can
now be associated with particular service classes and
IPv6 routing performed on the basis of these classes,
thus improving the performance of core routers. The
benefits of this scheme (Fgee, 2004) are its
simplicity and speed.
4 IPV6 QOS MANAGEMENT
The proposed QoS scheme handles QoS requests by
processing, monitoring, and controlling the traffic
flows. Weighted Fair Queuing (WFQ) is used to
separate the flows in which separate queues are
assigned for each traffic flow. The IPv6 QoS
management system uses the DGI and traffic class
(TC) field for reserving and tracking traffic flows.
This scheme is unique in that the IPv6 network can
be managed without invoking any other QoS
protocols such as RSVP or MPLS. The end-to-end
delay is less as the backbone routers use the DGI for
forwarding decisions as compared to the longest
match procedure used by other schemes. This
technique is scalable as the edge routers handle QoS
requests and communicate with other QoS
managers. Traffic flows are classified based on the
TC field so that each priority level is treated
differently. A sender sends its QoS request to the
network edge router. Upon receiving requests, the
edge router communicates with its domain manager
to approve or reject these requests. The edge router
forwards the manager’s responses, either positive or
negative, to the sender. When accepted, the source
starts sending data packets to the edge router where
traffic flow packets are classified, scheduled and
monitored. Packets are queued depending on their
TC field and the policies set by the QoS manager.
The leaky bucket algorithm is implemented to police
incoming traffic. The algorithm parameters for the
accepted traffic are setup according to their traffic
specifications. When a flow violates its requested
specification, its priority level is degraded or its
packets dropped.
5 DYNAMIC PRICING MODEL
The pricing model is based on the DiffServ end-to-
end pricing scheme introduced in (Tianshu, 2003).
In this market based model the value for the base
price
and fill factor is set by the network
provider for each traffic type. The fill factor is the
ratio of target capacity
to the maximum capacity
for a service class i. The price for a
class i at time t is computed (Tianshu, 2003, Waog,
2001) by:
i
base
P
i
f
i
T
i
C
max
)(tP
i
iiiiii
TTDtPtP /)()1()(
+
=
α
Here,
is the demand or current load for class i
and
i
D
i
α
is the convergence rate factor that determines
ICEIS 2005 - SOFTWARE AGENTS AND INTERNET COMPUTING
290
how the price converges to its maximum. Figure 1
illustrates the general pricing strategy. When the
load for a particular service class is lower than its
targeted capacity, the price is the base price
i
base
P
at particular service class. As the load exceeds
its target capacity, and when the load is close to the
maximum capacity, the price is increased rapidly i.e.
we have a dynamic pricing scheme where the price
is a function of current network conditions.
for th
Figure 1: General pricing strategy
During demand increase we adopt the exponential
pricing strategy.
P
=
otherwiseeP
TDif
tP
i
i
T
D
i
base
ii
i
base
i
]1[
)(
α
A price limit,
, indicates the price when the
s m
i
P
max
demand reache aximum capacity is set for each
service class that and is given by:
]1[
i
i
D
α
max
=
i
T
i
base
i
ePP
Knowing
d the fill factor for a
ss i ves th ve
i
P
max
,
i
base
P
an
i
f
service cla gi e solution for the con rgence
factor
i
α
that determines how the price converges
to the m ximum. a
)
1
(*)log(
max
i
i
i
base
i
P
i
f
f
P
=
α
The total revenue is the sum of all classes’ prices.
5.1 Pricing integrated QoS manager
Figure 2 shows the flow chart of the proposed
rou
6 SIMULATION RESULTS
We study the behavior of the integrated pricing
.04 pe
cenarios have been tested for each
pricing model. This figure summarizes the functions
performed by the various network entities in
implementing the pricing strategy. The source
initiates the QoS request and waits for responses that
include the acceptance messages and the associated
prices. The edge router forwards requests and
responses. It monitors all traffic packets entering the
domain. The IPv6 QoS manager processes the QoS
requests and then sends the network status for each
accepted traffic flow to the pricing agent. The
pricing agent calculates the price for each accepted
traffic flow by first finding
i
f using the
information received from the domain QoS manager
which includes the expected traffic rate (
i
T ) and the
max allowed rate (
i
C
). The pricing agent
attached to the edge ter initiates the price
according to the network status and the defined base
price (
i
base
P
) for each traffic class accepted by the
manager. Prices are sent to the customers that
initiated requests who accept or reject the prices.
max
model in an IPv6 QoS capable environment using
the ns-2 simulator (NS-2, 2004). Figure 3 illustrates
the network topology used in simulations and
consists of 4 core and 2 edge routers. The Ingress
router acts as the pricing agent and handles QoS
requests generated by source1 and source2 nodes.
The total capacity of each link is 1 Mpbs with a
propagation delay of 1 msec. The traffic flow
specifications are: Source 1 has a traffic rate of 500
Kbps with priority 15 and Flow ID-15. Source 2 has
a traffic rate of 250 Kbps with priority 12 and flow
ID 12. Source 3 has a traffic rate is 250 Kbps, is
classified as Best effort type with flow ID- 8. The
base prices
i
base
P
for each class is set at $0.16,
$0.09 and $0 r unit time respectively starting
with the flow with the highest priority. Simulation
was performed on three flows with FID-15 having
the highest priority when the total link capacity
reaches 50%, FID-12 has the 2
nd
highest priority
when the total link capacity reaches 70% and FID-8
is classified as Best Effort for all link capacities
(Tianshu, 2003).
Two simulation s
traffic flow, one when the total link load is less than
the percentage assigned for each flow and the other
one when the traffic exceeds these percentages.
Figure 4 shows the change of the prices for traffic
flow FID-15.
IMPLEMENTING A DYNAMIC PRICING SCHEME FOR QOS ENABLED IPV6 NETWORKS
291
Figure 2: Flow chart for the IPv6 pricing model
Figure 3: Simulated pricing model network
Figure 4: Prices for the three traffic flows FID(15) &
FID(12) and FID(8)
From Figure 4 it is seen that the prices change
rapidly as the load increases which results in more
revenue. We see that the changes in the prices for
traffic flows FID-12 does not change much since its
percentage is set to 70% and the packets of this flow
are not as critical as the first one. This also results in
a small increase in the revenue compared with the
first one. However, during congestion packets
belonging to FID-12 are degraded to best effort as
per the policies set by the QoS manager. Best Effort
flow packets have no change in prices since the
expected load is set to 100%, however, they are the
first to be dropped during congestion.
Source
1-Send QoS request
2- Check price &
resource availability
3- Send data &
disconnect
IPv6 QoS Manager
1-Receive QoS requests
2-Sends responses to
requests
3-Provides network stats
used to calculate prices
4- Calculate weights used to
define
per flow
max
i
C
Edge Router
1- Receive/forward QoS
request
2- Classify, schedule
incoming traffic
3- Monitor incoming traffic
4- Communicate with
domain
p
ricin
g
a
g
ent
7 CONCLUSION
Pricing Agent
1- Calculate
C , .
max
i
i
f
2- Updates prices
3- Calculate domain
revenue
The main objective in IP billing schemes is to price
network resources dynamically, especially during
congestion. In this paper, an IPv6 QoS manager was
integrated with a DiffServ dynamic pricing model.
The manager allows only higher priority traffic
packets that are willing to pay more to be processed
during congestion. In this approach the pricing is
decoupled from QoS decisions/reservations thus
avoiding per-flow based messaging for either pricing
or admission control. In addition to the scalability
and simplicity, more revenues are generated as
prices change dynamically according to the network
status.
REFERENCES
Fgee, E., Kenney J., Phillips, W.I., Robertson, W.,
Sivakumar, S., “Implementing an IPv6 QoS
management scheme using flow label & class of
service fields”, IEEE CCECE 2004. vol 2 , pp.851- 54.
S. Faizuullah and Marsic, I, “
Pricing QoS: Simulation and
Analysis”,
Proceedings. 34th Annual Simulation
Symposium, 2001, Pp 193-199.
Tianshu Li, Y. Iraqi and R. Boutaba, “Pricing and
admission control for QoS-enabled Internet",
Computer Networks, 2004 Published by Elsevier.
S. Jha and M. Hassen, “Engineering Internet QoS”, Artec
House Boston, London 2002.
Tianshu Li, Y. Iraqi and R. Boutaba, "Traffic-Based
Pricing and Admission Control for DiffServ
Networks", Integrated Network Management, 2003.
IFIP/IEEE 8
th
Intl. Symposium, pp. 73-86.
X. Waog and H. Schulzrine, "Pricing Network Resource
for Adaptive Applications in a Differentiated Service
Network", INFOCOM 2001., Vol 2 , Pp: 943 – 952.
NS-2 The Network Simulator ns-2, Oct. 30, 2004.
http://www.isi.edu/nsnam/ns/.
ICEIS 2005 - SOFTWARE AGENTS AND INTERNET COMPUTING
292