You are on page 1of 4

REALIZATION OF DYNAMIC ADDRESSING SCHEME IN

IPV6-BASED MULTI SERVICE AND MULTI PROVIDER


ACCESS NETWORKS
Krisztina Nagy, Béla Richárd Pallos, Csaba Lukovszki
{nagy@mail., pallos@mail., lukovszki@}tmit.bme.hu

1 Introduction
In the next coming decades there will be an ever ascendant need for flexible access with multi
service, multi provider (MSMP) support on the uniform manner in the access network. This
factor could establish the base for the better competition between access providers, giving the
possibility for flexible service delivery for customers. Providers such as Network Access
Providers (NAPs) and Network Service Providers (NSPs) are going to play important roles in
the access network. NAP provides connectivity and data forwarding roles through the access
network. The NSP provides real service connectivity towards the internet and it is
implemented by a special NSPs, the Internet Service Providers (ISPs) and Application
Providers (APs) which are situated at the service network that is connected to the access
network through the regional network. The two particular providers, providing services
through access and regional networks to the services of the Internet and APs, are the NAP and
NSP. In the MSMP environment customers can arbitrarily choose among NAPs and NSPs
that must operate on a shared physical medium.
In our paper we describe the proposed provisioning model for future access networks, giving
a better understanding of the IPv6 hierarchical addressing and its application in the access
networks. Later, our proposed flexible addressing allocation scheme will be described. First
we give a theoretical model and then we present its implementation with the Dynamic Host
Configuration Protocol (DHCP). Finally, we show the usability with a simulation.
1.1 Access architecture
Today’s computer networks consist of two main parts. The core network is transparent from
users and it guarantees big baud rate and capacity, while the other part assures connectivity to
this backbone network. The access part could be split to different parts as it was depicted in
Figure 1.

Internet backbone Services network

NSP AP
ISP
tree
ISP NSP
topology
Regional network NSP NSP
bipartite
Access network topology
EN EN
NAP
EN NAP
EN NAP
EN NAP
EN
bipartite
topology
AN AN AN

HG HG AN AN AN AN AN AN AN
HG

CPE HG HG HG HG HG HG HG
Home network

Figure 1 Access to the backbone network. Figure 2 Logical addressing architecture


1.2 The MSMP environment
Today, so more the average customer use more applications in the same time. Besides this, we
have the possibility to choose a provider. But it is not ensured, that these service packages are
the most optimal if we require it from one provider. To resolve this problem, the multi service
– multi provider (MSMP) environment was idealized. The MSMP opens the door to the
customer to have resort to any services from any providers. In an extreme case it means that
each service is furnished through different providers.
The access services are implemented in the NAPs. When a customer requires for a service, it
implicates taking some NAPs which could supply it. The customer can choose among the
NAPs and, similarly, among NSPs. As the result, in the ideal MSMP environment, each
customer application could arbitrary connected to any NAP and NSP pair.
The easiest way to manage this environment, addressing seems to be the optimal solution. If
we distinguish NAPs in virtue of address spaces, there can be no problem to claim for more
services in the same time. Implemented IPv6 stacks offer the possibility to handle more
addresses in one host so in the following we are going to examine in this way. In the proposed
model a customer has as many IPv6 addresses as many NAP-NSP pairs it is connected to. The
logical architecture of addressing in the access network for MSMP environment can be seen
in Figure 2.
2 Flexible Address Allocation Strategy
2.1 Motivation
Inspecting these prospects, we have proposed to work out a new model which can handle the
problems, we have described above. Accordingly it is hierarchical addressing model, which is
flexible, so it can manage incoming address demands in a dynamically changing access
network. The greatest significance of the model is prefix delegation that can be initiated from
any level in the hierarchy. To achieve this, we have improved our address aggregation and
delegation methods. Henceforward customers have the possibility to demand an address space
for each service they would like to resort. In this section we describe the main methods and
we are going to bring on the performance of the model, focusing on the major differences and
results.
2.2 Aggregation
In our model, the aggregation of the incoming demands is the principle. This means that a
node in the hierarchy aggregates the incoming prefix demands and will forward only one
aggregated request to its upper node. Of course this new value has to be as large as to serve all
requests.
The minimal number, that the address request to the given node could be served, can be
calculated from Eq. (1).
  N 
n = log 2  ∑ 2 ni  (1)
  i =1 
Where, N is the number of incoming request, and each request is represented with a bit
number of ni.
2.3 Delegation
Address delegation defines for a node the way to assign the available addresses to its lower
neighbours. Before all, we need to take some reservations. Recent routing models of the
Internet work with a network mask. [8] To adapt our method to this, we have to state that the
size of allocated addresses is the power of 2.
The speciality of AN is, it reserves a part of its address space for future use. This is a simply
practical restraint because topology variation is most significant on this layer. In a word,
avoiding common reconfiguration of the whole network, ANs are going to allocate as much
addresses as requested and not more. So they have spare addresses for later demands which
they can serve without renumbering.
2.4 Renumbering
Renumbering, initiated by the ISP. There are two reasons when the ISP has to start
renumbering. The first is if a new request can not be served from the actual allocation. In that
case, the ISP disposes of utilization information (for example size of the largest free address
space / whole address space). If some of these values overtake a limit, the ISP should
reconfigure the network from its address poll, reserved for renumbering.
3 Implementation with DHCPv6 PD
The Dynamic Host Configuration Protocol for IPv6 is a new improvement (see [2]). It was
designed for stateful auto configuration in a network. It is a client-server model, where hosts
can ask the server for network configuration parameters. Connecting to a new network,
DHCP-capable client automatically discovers the nearest server and sets its parameters
according to the given values. Server discovery can be written is four steps:
A new extension of DHCPv6 was set up. DHCPv6 with Prefix Delegation (see [4]) enables
asking for an address space, in a prefix format, instead of only one IPv6 address. With this, it
is possible to connect private domains to the network. Our model fits together with this
extension.
4 Simulation
4.1 The simulation environment
In the simulation and measurement environment we have made some simplification on the
model. The customers may choose one of the providers by requesting address pool from the
chosen provider. The customer may attend for multiple services with more IP addresses. In
the simulation we treated these customers as more than one client. This simplification does
not affect the measurement results about renumbering but with it is possible to treat the
hierarchical model as a tree graph.
The nodes periodically renew their leases. If the lease time is about to expire, the nodes try to
get another prefix for renumbering and start the renumbering procedure. This case the renew
requests originated from the lower entities will be rejected and the node offers a prefix from
the renumbering pool. In the simulation, the ISP node halves its address pool. Only one half
will be allocated for the NAP nodes according their used/unused pool size ratio, the second is
reserved for renumbering.
4.2 Numerical results
We have made the measurements for the first 800 time intervals from the initial state. On the
it can be seen that after this interval the pool is very close to be full. The depletion of the
address pool causes service failure for the newly arrived request.
In Figure 3 and Figure 4 we plotted the address space utilization at the same aggregation node
using static address space distribution and our dynamic addressing model with renumbering.
The comparison was made with the same behaviour at the customers’ side.
Figure 3. Address space depletion without Figure 4. Address space depletion with renumbering
renumbering using static addressing using dynamical addressing
In the static model the aggregation node manage the request from only its address space, since
the utilization reached almost the maximum during the simulation, while at the other
aggregation nodes remained low. In our model the address space always adapted to the actual
aggregated request size, so the utilization reached only 10 percent. About the time of 40 it
received 3 additional bits from the NAP address space to handle changing address requests.
5 Conclusion
It is straightforward from the results that our model always adapts to the aggregated requests
at any level of the hierarchy, in that way giving flexible addressing model for multi-service,
multi-provider access network architecture.
6 Zusammenfassung
Unser Artikel stellt eine IPv6-Addresszuweisungsarchitektur für multi-service, multi-provider Zugangsnetzwerke
vor, die es den Kunden erlaubt, beliebige Zugangsnetzwerke zu wählen (arbitrary mode). Dank ihrer dynamischen
Natur passt sich die Addresszuweisung automatisch an die Ansprüche der Benutzer sowie an die Änderungen der
Topologie an. In diesem Artikel werden die Zugangsumgebung, das verwendete Addresszuweisungsmodell, und seine
Anwendung im Rahmen von DHCPv6 PD beschreiben; außerdem stellen wir Simulationsergebnisse vor, die die
Brauchbarkeit der Architektur belegen.

7 References
[1] A. Inomata et all, “IPv6 Address Allocation and Assignment Policy”, RIPE 267, 22 January 2003
[2] R. Droms, J. Bound, B. Volz, T. Lemon, C. Perkins, M. Carney, “Dynamic Host Configuration
Protocol for IPv6 (DHCPv6)”, RFC 3315, IETF Network Working Group, July 2003
[3] S. Miyakawa, R. Droms, “Requirements forIPv6 Prefix Delegation”, RFC 3769, IETF Network
Working Group, June 2004
[4] O. Troan, R. Droms, “IPv6 Prefix Options for Dynamic Host Configuration Protocol (DHCP)
version 6”, RFC 3633, IETF Network Working Group, December 2003
[5] R, Hinden, S. Deering, “Internet Protocol Version 6 (IPv6) Addressing Architecture”, RFC
3513, IETF Network Working Group, April 2003
[6] R. Hinden, S. Deering, “Internet Protocol, Version 6 (IPv6) Specification”, RFC 2460, IETF
Network Working Group, December 1998
[7] “Guidelines for 64-bit Global Identifier (EUI-64) Registration Authority”, IEEE,
http://standards.ieee.org/regauth/oui/tutorials/EUI64.html, March 1997
[8] V. Fuller, T. Li, J. Yu, K. Varadhan, “Classless Inter-Domain Routing (CIDR): an Address
Assignment and Aggregation Strategy”, RFC 1519, IETF Network Working Group, Sept. 1993
[9] S. King et all., “The Case for IPv6”, IAB internet draft, 25 June 2000
[10] R. Hinden, M. O’Dell, S. Deering, “An IPv6 Aggregatable Global Unicast Address Format”,
RFC 3587, IETF Network Working Group, July 1998

You might also like