From falk at ISI.EDU Thu Feb 2 22:25:33 2006 From: falk at ISI.EDU (Aaron Falk) Date: Fri, 3 Feb 2006 15:25:33 +0900 Subject: [xcp] XCP paper at PFLDnet Message-ID: <2F29D77E-7817-4095-BA32-2A0EA980D9BB@ISI.EDU> Just heard this paper at PFLDnet 2006. XCP is useful to optical networks because it allows fair sharing under conditions of very expensive network buffering. --aaron Optical Rate-based Paced XCP for Small Buffered Optical Packet Switching Networks Abstract? One of the difficulties of OPS networks is buffering optical packets in the network. O(1) reading operation is not possible in the optical domain, because there is no equivalent optical RAM available for storing packets. Currently, the only solution that can be used for buffering in the optical domain is using long fiber lines called Fiber Delay Lines (FDL). FDLs provide a small and fixed amount of buffering. Burstiness of Internet traffic and short-term over-utilizations cause high packet drop rates in small and fixed buffered OPS networks. In this paper, we propose an architecture using a XCP- based congestion control algorithm specially designed for OPS WDM networks with pacing at edge nodes for minimizing the buffer requirements at core nodes. We show how the FDL requirements change with the number of flows on a wavelength, FDL granularity and wavelength utilization by using a dumbbell topology. We show the requirements and parameter settings for stable operation and high utilization. full paper: http://www.hpcc.jp/pfldnet2006/pdf/s6_01.pdf From fla at inescporto.pt Thu Feb 9 09:29:21 2006 From: fla at inescporto.pt (Filipe Abrantes) Date: Thu, 09 Feb 2006 17:29:21 +0000 Subject: [xcp] Header size Message-ID: <43EB7BF1.50108@inescporto.pt> Hello all, This may look like a stupid question, but why is the XCP header 40 bytes long? 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Protocol | Length |Version|Format | unused | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | RTT | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | X | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Delta_Throughput | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Reverse_Feedback | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Length: 8 bits This field indicates the length of the congestion header, measured in bytes. Length in this version of XCP will always be 40 bytes, or 0x28. I only see 20 bytes in the XCP header. Does it include TCP header? Does it include 2-way XCP headers? Im just clueless. Thank you and forgive me for my ignorance :), Filipe -- Filipe Lameiro Abrantes INESC Porto Campus da FEUP Rua Dr. Roberto Frias, 378 4200-465 Porto Portugal Phone: +351 22 209 4266 E-mail: fla at inescporto.pt From yuri at ISI.EDU Thu Feb 9 10:37:52 2006 From: yuri at ISI.EDU (Yuri Pradkin) Date: Thu, 9 Feb 2006 10:37:52 -0800 Subject: [xcp] Header size In-Reply-To: <43EB7BF1.50108@inescporto.pt> References: <43EB7BF1.50108@inescporto.pt> Message-ID: <200602091037.52504@kmail.yap.isi.edu> Thank you for pointing out a typo. We all must have subconsciously switched to 64-bit words to make a mistake like that :-) It should say "Length in this version of XCP will always be 20 bytes, or 0x14." In the implementation though this header could be followed by a "debug" (sub)header so the length could be longer (this behavior is controlled by a sysctl). -Yuri On Thursday 09 February 2006 09:29, Filipe Abrantes wrote: > Hello all, > > This may look like a stupid question, but why is the XCP header 40 > bytes long? > Length: 8 bits > > This field indicates the length of the congestion header, > measured in bytes. Length in this version of XCP will always be 40 > bytes, or 0x28. > > I only see 20 bytes in the XCP header.