23-07-2012, 12:28 PM
A proposed architecture for integrating Active Networks and MPLS
[attachment=28461]
INTRODUCTION
The main function of a network is to deliver packets from one
end-point to another. However, there are situations, mainly in
access areas, where packets need to be processed within the
network. This is the case of firewalls, Web proxies, multicast
routers, mobile routers or other similar services that need to
access information from a packet header to decide whether
the current packet should be dropped or how it should be
forwarded. More generally, the packet header or data may be
modified.
As MPLS evolves beyond the core network, penetrating
the access area, a big issue that arises is its inability to perform
packet computations for services such as the ones described
above. Active Networks is a novel solution for implementing
such services providing a flexible network infrastructure with
increased capabilities.
MPLS Overview
MPLS [1, 2, 3] is a packet label-based switching technique
which was originally devised to perform fast switching in
the core of the network. The argument for that was that the
algorithm for the “longest-prefix match” is more complex and
time consuming than the exact match used by MPLS, where
the label is an index into a table.
In time, MPLS proved to have other qualities even more
appealing than fast switching [4]. Being a relative simple
connection-oriented protocol, it proved to be suitable for
implementing traffic engineering and Quality of Service
(QoS) Routing in a simpler way than using IP.
Terminology in MPLS
The main elements of an MPLS network are the
MPLS nodes/routers which are called Label Switching
Routers(LSRs). They forward the packets within an MPLS
network. The routers at the edge of the MPLS domain are
called Label Edge Routers(LERs) and they examine the
packets headers and based on the information within those
headers the packet is assigned to a class called Forwarding
Equivalence Class(FEC). All packets within the same Forwarding
Equivalence Class will be labelled with the same
label and will be treated equally in the network.
The Active Nodes Architecture
The authors of [7, 8] refer to this as the discrete model or
Programmable switches approach because the programs and
data are carried separately (i.e., are discrete), while [9] refers
to it as the active node approach.
In this approach, the packets carry some identifiers or
references to predefined functions that reside in the active
nodes. The packets are active in the sense that they decide
which functions are going to be executed on their data, but the
actual code resides in the active node.
CONCLUSIONS
We propose in this article the integration of two edge technologies,
the Multiprotocol Label Switching and Active Networks.
Integrating these technologies overcomes a significant limitation
of MPLS in access networks, namely its inability to perform
switching above layer two. We have implemented a prototype
network using Linux, which proves the validity of the
concept. Future work will address how to exploit the flexibility
of the active MPLS concept to address tasks such as firewalling
without requiring inspection of the IP or TCP headers.
[attachment=28461]
INTRODUCTION
The main function of a network is to deliver packets from one
end-point to another. However, there are situations, mainly in
access areas, where packets need to be processed within the
network. This is the case of firewalls, Web proxies, multicast
routers, mobile routers or other similar services that need to
access information from a packet header to decide whether
the current packet should be dropped or how it should be
forwarded. More generally, the packet header or data may be
modified.
As MPLS evolves beyond the core network, penetrating
the access area, a big issue that arises is its inability to perform
packet computations for services such as the ones described
above. Active Networks is a novel solution for implementing
such services providing a flexible network infrastructure with
increased capabilities.
MPLS Overview
MPLS [1, 2, 3] is a packet label-based switching technique
which was originally devised to perform fast switching in
the core of the network. The argument for that was that the
algorithm for the “longest-prefix match” is more complex and
time consuming than the exact match used by MPLS, where
the label is an index into a table.
In time, MPLS proved to have other qualities even more
appealing than fast switching [4]. Being a relative simple
connection-oriented protocol, it proved to be suitable for
implementing traffic engineering and Quality of Service
(QoS) Routing in a simpler way than using IP.
Terminology in MPLS
The main elements of an MPLS network are the
MPLS nodes/routers which are called Label Switching
Routers(LSRs). They forward the packets within an MPLS
network. The routers at the edge of the MPLS domain are
called Label Edge Routers(LERs) and they examine the
packets headers and based on the information within those
headers the packet is assigned to a class called Forwarding
Equivalence Class(FEC). All packets within the same Forwarding
Equivalence Class will be labelled with the same
label and will be treated equally in the network.
The Active Nodes Architecture
The authors of [7, 8] refer to this as the discrete model or
Programmable switches approach because the programs and
data are carried separately (i.e., are discrete), while [9] refers
to it as the active node approach.
In this approach, the packets carry some identifiers or
references to predefined functions that reside in the active
nodes. The packets are active in the sense that they decide
which functions are going to be executed on their data, but the
actual code resides in the active node.
CONCLUSIONS
We propose in this article the integration of two edge technologies,
the Multiprotocol Label Switching and Active Networks.
Integrating these technologies overcomes a significant limitation
of MPLS in access networks, namely its inability to perform
switching above layer two. We have implemented a prototype
network using Linux, which proves the validity of the
concept. Future work will address how to exploit the flexibility
of the active MPLS concept to address tasks such as firewalling
without requiring inspection of the IP or TCP headers.