Proceedings of the International Conference , “Computational Systems and Communication
Technology” Jan.,9,2009 - by Lord Venkateshwaraa Engineering College,
Kanchipuram Dt.PIN-631 605,INDIA
Advanced Design and evaluation of a TIPP protocol for a New
Inter-Domain Routing Architecture (NIRA)
By
GOPIRAJAN.P.V, LECTURER, SAVEETHA ENGINEERING COLLEGE
ABSTRACT
In today’s Internet, users can choose
2. Related Work
their local Internet Service providers (ISPs), but once
Related work falls into three categories: Scalable
their packets have entered the network, they have little
routing schemes, routing architecture proposals,
control over the overall routes their packets take.
and current route selection technologies.
Giving a user the ability to choose between provider-
a) Scalable Routing:
level routes has the potential of fostering ISP
Scalable routing schemes aim to reduce the amount of
competition to offer enhanced service and improving
routing state a router keeps. The well-known schemes
end-to-end performance and reliability. This paper
include the cluster-based hierarchical routing, the
presents the design and evaluation of a new Internet
landmark hierarchical routing system, geographical
routing architecture (NIRA) that gives a user the ability
routing, and hybrid routing. However, the goals of
to choose the sequence of providers his packets take.
these routing systems are fundamentally different from
1. Introduction
us. We aim to provide a feasible and scalable approach
NIRA addresses a broad range of issues, including
to support user-controlled routes, while they aim to
practical
route
reduce the size of routing tables, or the number of
discovery, efficient route representation, fast route fail-
routing updates, and do not necessarily support user-
over, and security. NIRA supports user choice without
selected
running a global link-state routing protocol. It breaks
addressing has long been proposed to scale the
an end-to-end route into a sender part and a receiver
Internet routing. Our design builds on the idea of
part and uses address assignment to represent each
hierarchical addressing, but we developed a complete
part. A user can specify a route with only a source and
routing system that supports user route choice.
a destination address, and switch routes by switching
b)
addresses. We evaluate NIRA using a combination of
Nimrod
network measurement, simulation, and analysis. Our
mechanism for a source to discover network topology
evaluation shows that NIRA supports user choice with
and to use virtual circuit to set up routes. However,
low overhead.
Nimrod does not address how to fit its design into a
provider
compensation,
scalable
routes.
hierarchical
Routing Architecture Proposals:
proposes
Copy Right @CSE/IT/ECE/MCA-LVEC-2009
1
Provider-rooted
to
use
a
map-distribution
Proceedings of the International Conference , “Computational Systems and Communication Technology” Jan.,9,2009 - by
Lord Venkateshwaraa Engineering College,
Kanchipuram Dt.PIN-631 605,INDIA
policy-rich inter-domain routing environment. In
of NIRA is essentially different. NIRA aims to allow
contrast, our design is optimized to fit into the Internet
users to choose provider-level routes, and includes a
routing environment, and preserves the packet-
protocol TIPP that distributes routes and topology
switched feature of the Internet. The Inter-domain
information to users for them to choose routes. We
Policy Routing (IDPR) protocol proposes to use a
note that NIRA and the IPv6 site multi-homing
domain-level link state routing protocol to distribute
proposal share some similarity. The IPv6 proposal
the domain-level topology to route servers of each
also gives multiple addresses to a multi-homed site. A
domain. A source sends a route request to a route
key difference between NIRA and the IPv6 proposal
server to obtain a domain-level route. The Scalable
is that NIRA considers the address selection problem
Inter-Domain Routing Architecture proposes to use a
as a path selection problem. It provides necessary
BGP router’s routing information base and to flood a
topology information for users to select a path. The
route request from a destination to a source to assist a
protocol TIPP propagates topology information
user to discover multiple routes from the destination to
associated with addresses to users. This allows a user
the source. NIRA is designed to handle IP layer
to map an address to a provider-level route, and to
routing and addressing issues. The design goal is
choose an initial source address that is failure-free.
fundamentally different. This includes a Wide-Area
Moreover, NIRA uses both the source and the
Relaying Protocol (WRAP) that provides extended
destination address to forward a packet. This allows a
addressing and improved loose-source routing. NIRA
user to control the domain-level route, including both
follows the good lead of WRAP to use the path-based
the part of the route in the sender’s access network,
addressing scheme. Feedback Based Routing proposes
and the part of the route in the destination’s access
to use a domain-level link-state routing protocol for
network. In contrast, the IPv6 proposal does not
edge routers to learn domain-level topology. An edge
change the routing paradigm of the Internet. Routes
router selects a route to reach a destination, monitors
are chosen by routers, and forwarding is destination-
the route condition, and switches to a different routes
based. As a result, the source address of a packet
if the route fails. In contrast, our work does not require
influences the return path a packet takes, but do not
global link-state routing. We can leverage the route
determine the outgoing path the packet follows. An
monitoring algorithm described in this proposal for
end host finds a working path by exploring address
rapid route fail-over. The HLP proposal intends to
pairs without knowing the providers the addresses
improve the scalability of BGP. HLP uses a mixture of
map to. Further, NIRA allows a user to choose
a link-state routing protocol and a path vector routing
beyond the first hop provider.
protocol to provide fast routing convergence. The goal
c) Current Route Control Technologies:
Copy Right @CSE/IT/ECE/MCA-LVEC-2009
2
Proceedings of the International Conference , “Computational Systems and Communication Technology” Jan.,9,2009 - by
Lord Venkateshwaraa Engineering College,
Kanchipuram Dt.PIN-631 605,INDIA
Both commercial route control products and overlay
necessarily support user-selected routes. Provider-
networks offer route selection service to some extent.
rooted hierarchical addressing has long been pro-
Route control products are limited to selecting the next
posed to scale the Internet routing.
hop provider for outbound traffic, and cannot choose
ii) Routing mechanisms
beyond the first hop provider. Moreover, they are
Inter-domain Policy Routing (IDPR) A domain-
generally not affordable by individual users or small
level link state routing protocol to distribute the
sites. An overlay network has a limited scope. Only
domain-level topology to route servers of each
nodes on an overlay network can control their paths by
domain. A source sends a route request to a route
tunneling traffic through other nodes on the overlay
server to obtain a domain-level route.
network. Our work aims at providing a long term
solution to support user route selection. We introduce
IPV6
changes at the network layer, and once deployed, all
The IPv6 proposal does not change the routing
Internet users are able to benefit from our design.
paradigm of the Internet. Routes are chosen by
Today, users can pick their own ISPs, but once their
routers, and forwarding is destination-based. As a
packets have entered the network, the users have no
result, the source addresses of a packet influences the
control over the overall routes their packets take. ISPs
return path a packet takes, but does not determine the
make
outgoing path the packet follows.
business
decisions
to
interconnect,
and
technically the BGP routing protocol is used to select
5. Proposed system
the specific route a packet follows. Here we compare
all the already available methods, protocols and also
Here we differentiate our NIRA with the available
various algorithms for this NIRA.
methods and mechanisms.
4. Existing system
Over other Routing methods When comparing to
other routing methods, we aim to provide a feasible
i) Routing methods
and scalable approach to support user-controlled
The well-known schemes include the cluster-based
routes, while they aim to reduce the size of routing
hierarchical routing, the landmark hierarchical
tables, or the number of routing updates, and do not
routing system, geographical routing, and hybrid
necessarily support user-selected routes. Provider-
routing. In this type of routings it is not feasible and
rooted hierarchical addressing has long been pro-
scalable approach to support user-controlled routes,
posed to scale the Internet routing. Our design builds
while they aim to reduce the size of routing tables, or
on the idea of hierarchical addressing, but we
the number of routing updates, and do not
Copy Right @CSE/IT/ECE/MCA-LVEC-2009
3
Proceedings of the International Conference , “Computational Systems and Communication Technology” Jan.,9,2009 - by
Lord Venkateshwaraa Engineering College,
Kanchipuram Dt.PIN-631 605,INDIA
developed a complete routing system that supports
We motivate the key components of NIRA. The
user route choice.
overall goal of NIRA is to support user choice. To
achieve this goal, the design of NIRA must satisfy
Inter-domain Policy Routing (IDPR)
several implied requirements. First, to be practical,
Our work does not need a global link state routing
we require that NIRA must recognize the need for
protocol for a user to discover routes, and does not
payment and allow practical payment schemes. In a
require the presence of a per-domain route server.
commercialized
Internet
that
involves
multiple
IPV6
parties, an architecture design must respect users’
NIRA considers the address selection problem as a
economic interests as well as those of ISPs. A
path selection problem. It provides necessary topology
technical design will not be feasible if there is no
information for users to select a path. The protocol
practical mechanism for users to pay ISPs for their
TIPP propagates topology information associated with
service. So in our design, we do not require new
addresses to users. This allows a user to map an
payment schemes such as micro-payments. Second,
address to a provider-level route, and to choose an
we require that our design be highly efficient and
initial source address that is failure-free. Moreover,
scale well. A future Internet must support a wide
NIRA uses both the source and the destination address
range of devices, ranging from resource-constrained
to forward a packet. This allows a user to control the
sensor nodes to super computers, and a wide range of
domain-level route, including both the part of the route
communication technologies, ranging from low
in the sender’s access network, and the part of the
bandwidth wireless connections to fiber optics. An
route in the destination’s access network.
efficient and scalable design would accommodate this
variety and allow resource-constrained end systems to
Over other Route control technologies
take advantage of the design without additional
Our work aims at providing a long term solution to
infrastructure support.
support user route selection. We introduce changes
i) Design of TIPP protocol with route discovery.
at the network layer, and once deployed, all Internet
TOPOLOGY INFORMATION
users are able to benefit from our design. NIRA
PROPAGATION PROTOCOL (TIPP)
provides protocols to inform a user of the available
TIPP is an inter-domain protocol that runs between
routes so that provider selections can be done by
border routers of domains. It operates outside the
software at a fine granularity, such as at the
Core of the Internet, and has three functionalities:
granularity of per connection or per application.
1) TIPP automatically propagates the mapping
6. System Design Specification
between addresses
Copy Right @CSE/IT/ECE/MCA-LVEC-2009
4
Proceedings of the International Conference , “Computational Systems and Communication Technology” Jan.,9,2009 - by
Lord Venkateshwaraa Engineering College,
Kanchipuram Dt.PIN-631 605,INDIA
2) TIPP propagates up-graphs to users;
domain-level peering link the domain has with a
3) TIPP helps routers to establish inter-domain
peer. Therefore, those providers or peers may not
use the peering connection. The main idea of Tipp’s
forwarding entries.
We design TIPP to use separate protocol messages
topology update algorithm is that a node updates its
to propagate address information and topology
topology database using messages heard from its
information. This design choice minimizes the
neighbors,
coupling between the two functionalities. The part of
messages from different neighbors by “believing”
TIPP
is
the neighbor that is on the shortest failure-free path
straightforward: if a domain receives an address prefix
to the link that triggers the update messages.
allocation (or withdrawal) message from a provider, it
Messages sent along the same failure-free path are
sends a message to allocate (or withdraw) a
in-order and reliable, because messages sent
subdivision of the prefix to a customer. Each domain
between adjacent nodes are in-order and reliable.
also appends its own domain identifier in the message
Therefore, the sequences of messages coming from
to provide the mapping between an address and a route
the neighbor on the shortest failure-free path reflect
segment. This part is similar to a path-vector protocol.
the sequential status change of that link. A failure-
The part of TIPP that distributes topology
free shortest path can be recursively computed,
information is a policy-controlled link-state protocol.
because a node knows whether its adjacent link to a
A domain can control what to distribute to a
neighbor is failure-free, and the neighbors in turn
neighbor at the granularity of a link. A domain has
know which of their adjacent links are failure-free.
two types of control: scope enforcement and
Recursively, the node can determine a failure-free
information hiding. Scope enforcement ensures the
path to a link at any distance.
that
propagates
address
information
scalability of TIPP. A domain can choose not to
and
resolves
inconsistent
update
a). Initialize nira_ agent. Initialize and configure
propagate anything heard from a customer to its
out TIBs.
neighbors so that a link-state message will only be
b). Start the listening socket;
sent downward a provider hierarchy. Information
c). Start all configured peers.
hiding supports policy routing. A domain may only
d). Start aging timer. Aging for peer's in address
send a link-state message received from one
databases. After we synchronized our address
neighbor to another neighbor if it provides transit
Databases with our peer, we then need to let our
service between the two neighbors. With this policy,
peer know what address prefixes we'll announce to
the domain’s providers or other peers will not
it, and the set of reachable address prefixes if it
receive the link-state messages regarding to a private
sends us a packet. The peer needs the first piece of
Copy Right @CSE/IT/ECE/MCA-LVEC-2009
5
Proceedings of the International Conference , “Computational Systems and Communication Technology” Jan.,9,2009 - by
Lord Venkateshwaraa Engineering College,
Kanchipuram Dt.PIN-631 605,INDIA
information to build its forwarding table, and the
NRRS (Name-To-Route-Resolution Service) to
second piece for path selection. We store this
query for the route. If there is a cache hit, we fill in
information in the edgeRec_ structure of the peer.
the Nira header in the raw packet with the route in
The peer will propagate it to its neighbors depending
the route cache. Next, we need to find the interface
on its propagation policy. We also need to send the
to reach the next Hop in the route structure.
corresponding OutGrpah in outTIBs_ to the peer.
Pick a pair of source and destination addresses
Tipp's propagation policy: There are two types of
to compose routes, and remember our picks. If
policy:
unwanted_.size () == 0, pick a source and a
global and up Tree: Global policy means a
Tipp topology message will be propagated globally.
destination address that shares the longest prefix; if
Up tree policy means a Tipp topology message will
unwanted_.size () == 1, pick a source and a
be propagated within its provider tree. A node's non-
destination address that are maximally different from
hierarchical address is constructed from its region id
the ones in the unwanted_route; if unwanted_.size ()
and its own id. The first two bytes of a non-
> 1, pick a random source and a random destination
hierarchical address is the region id of a node. The
address that have not been picked before. If we've
last four bytes of a node's non-hierarchical address is
tried all source and destination combinations, we still
the node's own id. The ten bytes in the middle are set
cannot find a route. We return false.
to 0.
If the app does not tell us any unwanted routes, we
pick the pair of source and destination addresses that
share the longest prefix; if the app has told us one
unwanted route, we pick a pair of source and
destination addresses that are maximally different
from those in the unwanted route; if the app has told
us more than one unwanted route, we random pick a
source and a destination address. Note that for the
first pick, we can always pick the shortest source path
route.
Fig5.2 provider rooted hierarchical network
If the destination has prefix match in the
ii) Design of NIRA agent which forwards the
announced prefix list from the finish node to the start
packets along the failure free route.
node, we've found a path. If the path is not an
We first look up the route to reach the IP destination.
unwanted path, return.
If there is a route cache miss, we should invoke
Performance of the Protocol.
Copy Right @CSE/IT/ECE/MCA-LVEC-2009
6
Proceedings of the International Conference , “Computational Systems and Communication Technology” Jan.,9,2009 - by
Lord Venkateshwaraa Engineering College,
Kanchipuram Dt.PIN-631 605,INDIA
We show the average Message and Bytes, convergence
address to represent each part. A user can choose
time, etc. Our overall goal is to design a routing
routes by choosing addresses. As both the source
system that practically supports user choice. In this
address and the destination address are used for
section, we evaluate the technical aspects of NIRA that
forwarding, packets with arbitrary spoofed source
impacts its feasibility: scalability and efficiency. How
addresses will be dropped, and will not be forwarded
our design satisfies other practical constraints such as
to their destinations. NIRA includes a protocol TIPP
provider compensation and incremental deployment
that propagates to a user his addresses and the
are discussed in Section II. We evaluate the scalability
topology information associated with his addresses.
of NIRA’s route discovery mechanism. The evaluation
Our evaluation suggests that NIRA is practical. It
includes the amount of state a user obtains from TIPP,
supports user route choice with low overhead.
the overhead to
maintain the state, and the
References
convergence speed of TIPP. Our evaluation on
1. X. Yang. The Design and Evaluation of a
New Internet Routing Architecture (NIRA).
PhD thesis, M.I.T., Sept. 2004. MIT-LCSTR-967.
2. X. Yang. NIRA: A new Internet routing
architecture. In Proc.of ACM SIGCOMM
3. M. Steenstrup. An Architecture for InterDomain Policy Routing. IETF, June 1993.
Proposed Standard, RFC 1478.
4. D. Estrin, Y. Rekhter, and S. Hotz. Scalable
Inter-Domain Routing Architecture. In ACM
SIGCOMM, 1992.
5. S. Savage, T. Anderson, A. Aggarwal, D.
Becker, N. Cardwell, A. Collins,
6. E. Hoffman, J. Snell, A. Vahdat, G. Voelker,
and J. Zahorjan. Detour: Informed Internet
Routing and Transport. IEEE Micro, 1999.
7. “A Hierarchical Routing Model for Large
Scale
Networks
Based
on
Ant
Algorithm”Guozhen Tan Yi Liu Hengwei
Yao
Jialin Li
NingningHan Dalian
University
of
Technology,
116023,
P.R.China;
efficiency focuses on the connection setup latency
incurred by the reactive failure detection mechanism.
This setup latency affects the performance of
interactive applications and short transfers.
Conclusion and Enhancement
Giving a user the ability to choose domain-level routes
has the potential of fostering ISP competition to offer
enhanced
service
and
improving
end-to-end
performance and reliability. We present the design of
NIRA, an inter-domain routing system that practically
supports user choice. The design of NIRA addresses a
broad range of issues, including ISP compensation,
scalable route discovery, efficient route representation,
fast route fail-over, and security. NIRA supports user
route choice without running a global link-state
routing protocol. Our design splits an end-to-end route
into a sender part and a receiver part, and uses an
Copy Right @CSE/IT/ECE/MCA-LVEC-2009
7