GMPLS basic understanding
GMPLS can be used for traffic engineering, to create specific paths to transport flows of data.
GMPLS originates from MPLS, which is uses labels. While the technologies are quite different, it is still apparent if you look at the terminology.
GMPLS questions
- How exactly are destination or routers in a domain announced to other domains. I assume that each router (LSR and LER) get an IP address in a assigned range, and that those IP addresses are announced to neighbours, similarly to how BGP announces destination IP addresses.
- In a multi-domain scenario, it seem to use topology information (by announcing the IP addresses). However, it does not seem to announce availability of links (scheduled usage) or access permissions. This makes me wonder how path discovery is done, and which algorithm is used. Apparently, GMPLS can use constraint based routing. At least, the DRAGON project did look into CSPF PCE (Constraint-based Shortest Path First Path Computation Element), which is described in draft-ietf-pce-architecture∞.
- GMPLS has been extended to support multiple layers, like link, wavelength (WDM), time (TDM) and packet. However, I never saw a mention of Ethernet VLAN support. Perhaps this is not supported (maybe because ethernet uses statistical multiplexing rather then QoS traffic engineering?) or I just missed it, or that extension has just now been written (yet). The question in short is if one can use GMPLS if the LSR (label switch routers) are Ethernet switches with VLAN capacity.
- How does GMPLS handle technology transitions? For example, with WDM, let's assume there are two paths, one which has a red and blue channel available, and the other path with a green channel available. With WDM, some devices (LSC LSR -lambda switch capable label switch router- in GMPLS terminology) support wavelength conversions, while others don't. How does GMPLS handle this capability (and simular capabilities in other technology layers like TDM) in the routing algorithm?
- How to specify the encoding type for a link with IEEE 802.1q VLAN tags; how to distinguish between WAN PHY and LAN PHY encoding? What value should be used "LSP encoding type", "switching" or "Generalized PID"? I asume "LSP encoding type". WAN and LAN PHY contain both Ethernet. That seems obvious, but the question is more about the lower layer. This is in particular relevant with so-called UNI PHY, where the interface can switch between WAN PHY and LAN PHY.
- draft-dimitri-gels-framework-00 seems to suggest that perhaps new encoding types must be invented
- draft-papadimitriou-enhanced-lsps-04 explictly says that WAN PHY and LAN PHY fall in the Ethernet "LSP Encoding Type".
Categories
CategoryGMPLS
There are no comments on this page. [Add comment]