Design and implementation of a flexible, QoS-aware IP/ATM adaptation module

The overlaying of IP-based networks onto ATM subnetworks is a network configuration pattern found increasingly often. While IP networks traditionally only offer plain `best-effort' service they are now evolving to offer more sophisticated services. Nevertheless, the exact mechanisms for providing QoS in IP networks are not yet settled and essentially non-existing in today's production-level networks, with the Internet being the most popular and important example. On the other hand, ATM networks have been designed from their inception to offer a wide range of QoS mechanisms. Thus, given the configuration of an IP overlay network over an ATM subnetwork, it is very attractive to leverage ATM's QoS mechanisms to alleviate IP's QoS problem, at least partially. The invocation of those mechanisms will be done on so-called IP/ATM edge devices which are exactly at the frontier between the IP and ATM network. In this paper we describe the design and implementation of a flexible, QoS-aware IP/ATM adaptation module. This adaptation module allows an IP/ATM edge device to forward IP datagrams depending on their (header) contents onto specifically set up VCs in a performant manner. To achieve performance, it is necessary to implement this module in kernel space, at least partially. On the other hand, it should be easy to use, for e.g., an RSVP/IntServ over ATM, or a DiffServ over ATM mapping module. Therefore, the adaptation module is split into two parts, a kernel-level part that handles all the time-critical tasks of data forwarding and a user-level part which gives access to the functionality provided by the adaptation module.