smyers@miranda.dpl.scg.hac.com (Steven Myers) (02/12/91)
I have a question about how one would go about running the OSI ConnectionLess Network Protocol on X.25. In particular I'm interested in how X.25 virtual circuits are managed, and how routing and addressing is done. I would very much like to hear about actual router products and how their vendors accomplished this task. As I understand the problem, there are two different ways to go about it. 1) The first way is to open up X.25 Permanent Virtual Circuits (PVC) between your various OSI Intermediate Systems (IS). Under this scheme when ever that IS has packets to send to some other IS it just puts them out on the already open X.25 virtual circuit. The routing could be just a matter of mapping the OSI IS address to the proper X.121 addresses and determining which X.25 PVC goes to that X.121 address. The primary problem I see with this method is that it is very expensive to have a bunch of X.25 PVC's open. 2) The second method I see is for there to be some sort of protocol that allows OSI TP4 to open and close X.25 virtual circuits on the fly. To keep things simple, I'll assume that routing and addressing are done the same way as in 1). The big advantage with this method is that X.25 PVC's aren't needed, and thus your X.25 bill should be lower. In fact, it may even be possible to, as needed, open up more then one X.25 virtual circuit to improve your performance under peak load conditions. The primary problem I see with this method is the delay needed to open X.25 circuits, and how to determine when to close them. Maybe someone who knows more about X.25 can comment on how long it takes to open a virtual circuit on a typical X.25 network. I have heard that cisco does have a OSI-X.25 router that operates CLNP. I wasn't able to determine though which of the above methods is used. If they use method 2) how does their router determines when to close the X.25 circuit? If anyone has any product/vendor information they would like to send to me I will summarize and post to this news group. I will also summarize any comments on methods/protocols I receive. Thank you for your help. ------------------------------------+---------------------------------------- Steven Myers | Internet: smyers@luna.dpl.scg.hac.com Hughes | Phone: 213-414-6261 Space and Communications Group | FAX: 213-414-6699 Bldg. S64, MS C409 +---------------------------------------- PO Box 92919 | In UNIX anything is possible Los Angeles, CA 90009 | but nothing of interest is easy ------------------------------------+----------------------------------------
eskovgaa@CUE.BC.CA (Erik Skovgaard) (02/14/91)
You should probably look at a standard named "The Internal Organization of the Network Layer" (IONL) ISO 8648. When you run CLNS over X.25, you can view the Network Layer as being composed of three "sublayers": The Subnetwork Independent Convergence Protocol (SNICP), the Subnetwork Convergence Function (SNDCF) and the Subnetwork Access protocol (SNAcP). SNICP in you case is CLNP and SNAcP is X.25 (PLP). The function of SNDCF is to map the Network Address to the Subnetwork Access Point of Attachment (SNPA) address and in this case, to manage X.25 VCs. You can now open a VC when an IP packet is sent out. You do not have to close the VC, but can keep it open for the next IP packet. A timer takes care of the closing the VC if no traffic is taking place on the VC for a while. Otherwise it remains open. It would be too much overhead to open and close a VC for each IP packet. SNDCF is a local function and does not generate PDUs on its own. Other combinations of SNICP/SNAcP may require the use of a convergence protocol that generated PDUs. In that case, the middle sublayer is referred to as SNDCP (p - for Protocol). This is a gross simplification of IONL and SNDCF, but it should give you the general idea. Using VC's as opposed to PVCs allows you flexibility in routing. Have fun! ....Erik. ------------------------------------------- Erik Skovgaard PSC (Pacific) Inc. Vancouver, B.C.