Re: [Pce] RE: Comment on draft-ietf-pce-pcecp-interarea-reqs-00.txt

[JP Vasseur]

Hi, On Jan 13, 2006, at 4:22 AM, LE ROUX Jean-Louis RD-CORE-LAN wrote: Hi Zhang   sorry for replying so late...   Please see inline, De : Zhang Renhai [mailto:[EMAIL PROTECTED]] Envoyé : mercredi 28 décembre 2005 08:04 À : LE ROUX Jean-Louis RD-CORE-LAN Cc : [EMAIL PROTECTED] Objet : Comment on draft-ietf-pce-pcecp-interarea-reqs-00.txt Hi, Jean-Louis    I have some comment on the draft draft-ietf-pce-pcecp-interarea-reqs-00.txt    In section 7.11.1, In case of network failure, jittering will be used to avoid simultaneous requests sent to one PCE. Could more consideration be given here to the preemptment, becouse the jittering timeout is stochastic, some lower request may be served before a higher request and the path may be calculated differently. which may increase the probability of a preemptment.    JLLR: You raised a good point. First of all note that this point is not a protocol requirement, this is a local implementation issue, and the jittering was just given as an example... The PCC could apply different jitter to low priority and high priority requests, and could send high priority request before low priority one, this is a local PCC decision. Also the request prioritization function could be used so that the PCE serves high priority requests first. The way a PCE is going to handle request prioritites is also a local implementation issue. Maybe we should remove this sentence to avoid any confusion. By the way note that this section is generic and has just been moved to the generic requirement draft. So we will have to clarify in the generic draft.  Thanks indeed to remove this section since this is out of scope. JP.      I have always been thinking a question: if a PCC will not perform the CSPF  computation, why does it still maintain the TEDB any longer? which may consume a lot of memory and CPU of a LSR.This question dost not aid at this draft.    [JLLR] A LSR may be a PCC for some LSPs and a PCE for other LSPs.  Anyway a LSR that runs ISIS or OSPF for IP routing has to maintain a LSDB. I agree removing TE info would save some memory and CPU but the gain sounds quite low compared to the impact: For any computation, even a simple intra-area CSPF, you would have to ask a PCE, and this may lead to a high PCE stress. You would have a small CPU gain on LSRs at a cost of a huge CPU consumption on PCEs...   In inter-area environment,sometimes, a PCC may wish to get as many paths as possible, for all kinds of purpose,so could the PCC send the request to more than one PCEs?     [JLLR] Yes, and this is a procedure local to the PCC, that does not imply specific PCECP procedure.   Thanks for these valuable comments   Regards,   JL     Regards, Zhang _______________________________________________ Pce mailing list Pce@lists.ietf.org https://www1.ietf.org/mailman/listinfo/pce

_______________________________________________
Pce mailing list
Pce@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/pce