From saujain at neo.tamu.edu Thu Jan 5 07:26:05 2006 From: saujain at neo.tamu.edu (Jain, Saurabh) Date: Thu, 5 Jan 2006 15:26:05 -0000 Subject: [xcp] Error handling in XCP code (nodiv version) Message-ID: <200601051526.k05FQ5iT065065@xyzzy-3.tamu.edu> Hi All, How is error recovery done by the end hosts in the XCP code (nodiv version) whenever a packet is lost? Is the congestion window reduced to half and fast recovery and fast retransmission invoked while keeping the congestion window frozen as long as the host recovers from the error? Does the code uses SACK or can it be enabled? Thanks in advance, Saurabh From falk at ISI.EDU Mon Jan 9 10:45:23 2006 From: falk at ISI.EDU (Aaron Falk) Date: Mon, 9 Jan 2006 10:45:23 -0800 Subject: [xcp] Error handling in XCP code (nodiv version) In-Reply-To: <200601051526.k05FQ5iT065065@xyzzy-3.tamu.edu> References: <200601051526.k05FQ5iT065065@xyzzy-3.tamu.edu> Message-ID: <4CB8335E-B8F0-4A08-BE6A-94D148606571@isi.edu> On Jan 5, 2006, at 7:26 AM, Jain, Saurabh wrote: > Hi All, > > How is error recovery done by the end hosts in the XCP code (nodiv > version) > whenever a packet is lost? Is the congestion window reduced to half > and fast > recovery and fast retransmission invoked while keeping the > congestion window > frozen as long as the host recovers from the error? Does the code > uses SACK > or can it be enabled? > Saurabh- The error recovery code in the nodiv version is unchanged from the normal BSD TCP code. SACK is enabled. When TCP notices a loss (after sufficient DUPACKs have been received or an ECN mark, etc) it cuts cwnd in half. However, the cwnd growth is determined by the XCP algorithm, not slow-start and congestion avoidance. So, the next XCP packet arriving modifies cwnd by the network allocation. Unless the packet loss is from a congested XCP router (very unlikely), cwnd will rapidly grow -- in the same way as a new flow entering the network. This is not a intentional behavior. It is just a side effect of how we put the code in. If you run XCP in a network with all XCP routers and mostly lossless links, it is rarely executed. However, it is likely that any realistic deployment of XCP will be in a network with queues that are not XCP-enabled. We are currently developing new behavior for the scenario where the congestion bottleneck is a non-XCP router and packets are lost. More about this as it matures... Hope this helps, --aaron -------------- next part -------------- A non-text attachment was scrubbed... Name: PGP.sig Type: application/pgp-signature Size: 186 bytes Desc: This is a digitally signed message part Url : http://mailman.isi.edu/pipermail/xcp/attachments/20060109/c17e639d/PGP.bin From haldar at ISI.EDU Mon Jan 9 10:50:57 2006 From: haldar at ISI.EDU (Padmaparna Haldar) Date: Mon, 9 Jan 2006 10:50:57 -0800 (PST) Subject: [xcp] error with xcp-1.1 and ns2.28 In-Reply-To: References: <20051115160939.M60276@ee.ucl.ac.uk> Message-ID: Hi Moses, What kind of problem do you see when you run xcp_test.tcl? Are you trying to use nam? Thanks, --Padma >> From: "Moses Woldeselassie" >> Date: November 15, 2005 8:13:36 AM PST >> To: xcp at ISI.EDU >> Subject: [xcp] error with xcp-1.1 and ns2.28 >> >> Hi all >> >> I did try to use the XCP-1.1 and xcp_test.tcl with the ns2.28: >> >> I did replace these files queue.c and queue.h in nam-1.11. >> >> but my problem is with the xcp: xcp.cc, xcp-end-sys.cc and xcp-end-sys.h >> >> how do I modifie this Makefile/Makefile.in for the xcp.cc and xcp- >> end-sys.cc? >> what do I need beside that to run the xcp_test.tcl? >> >> I am looking forward to hearing from you. >> >> Cheers >> Moses >> >> *************************************************************** >> Electronic & Electrical Engineering @ UCL >> www.ee.ucl.ac.uk >> *************************************************************** >> >> _______________________________________________ >> xcp mailing list >> xcp at isi.edu >> http://mailman.isi.edu/mailman/listinfo/xcp