Different types of VPN

Our QoS model allows three types of VPNs as describedbelow.Type 1-static VPNs: For this kind of VPN, some preallocated dedicated or shared lightpaths are established tomeet its traffic needs. Traffic needing this type of servicenormally has a constant bit rate and require high servicequality, in terms of end-to-end delay and jitter. Staticlightpaths are for CBR traffic streams only and not forVBR or UBR/ABR traffic streams, even if there is no CBRtraffic on the static lightpaths. This will be the mostexpensive service due to the static lightpaths. Of course,within one VPN traffic matrix, there will be only a fewtraffic demands needing this type of service. Since there maybe several VPN traffic streams sharing the static lightpaths,it is necessary to have a scheduling mechanism in the buffersat the access nodes. A sophisticated scheduling mechanism,such as weighted fair queue can be adopted in thisframework. However, to simply illustrate the framework,we adopt a simple round-robin scheme here. In the nonbuffer scenario, a first come first serve principle is adopted.Type 2-OBS VPNs: For this kind of VPN, a set oflightpaths are provided that are established only whentraffic streams have the connection requirement. This kindof traffic stream is more likely to be generated by webbrowsers, wide area TCP connections and variable-bit-rate(VBR) video sources. These traffic streams are bursty.Lightpaths are established using optical burst switching(OBS). OBS sends a control packet first on a separatechannel, then waits for an offset time called just enoughtime (JET) [8], than the burst data packet is sent out. Afterthe burst data packet passes the link, the wavelengths arereleased.Type 3-alternative lightpath VPNs: For this kind of VPN, adynamic lightpath is setup when there is a traffic connectionrequirement. The lightpath setup uses a two-way reservationmethod, which introduces a long propagation delay wheninitialising the lightpath. Because of the delays involved inidentifying a path and is acknowledgement, this method issuitable for VPNs which have less stringent QoS requirements, e.g. UBR or ABR traffic streams.In this paper we study the performance of VPNs ofdifferent types. We examine the performance under twoscenarios: with buffers and without buffers in the accessnodes. In the scenario with buffers, we examine the delayfor the different types of VPN traffic streams. In the nonbuffer scenario we examine the blocking probability for thedifferent types of VPN traffic streams.For the static lightpath, we need a long-term trafficpattern and use a heuristic algorithm to set up the staticlightpaths. For the OBS lightpath, we follow a first comefirst service procedure to set up the lightpath for them. Ifthere are more than two traffic connection requirements atthe same time, the one with the heavier traffic loading is setup first. In the OBS lightpath, the data packets use only thedeterministic path to transfer. For the alternative lightpath,we use a two-way reservation scheme to set up the lightpathwhen the traffic connection requirement occurs. We seek analternative path if the deterministic path is not available.The latter two types of VPN can handle dynamic trafficstreams. Depending on the different QoS requirements, wedeploy OBS or the alternative scheme for different trafficstreams.