The program is acquiring a IP multicast flow on a thread and distributing all data received to 12 other processes on a socket connection (for a total of 12, one per process). The program is much slower on Fedora (processes about 2/3 of packets per second, compared to the one running on RH AS3.0) and uses more CPU as reported below. I was expecting that if this was a 2.6.5 kernel problem on Fedora, some of you could have already experienced the performance difference. The other explanation is that there is some kernel configuration I am missing. The system itself is a HP Proliant DL360G3, 2 3,0GHz processors, 4GB RAM. Regards Raffaele -----Messaggio originale----- Da: Carles Xavier Munyoz Baldó [mailto:carles@xxxxxxxxxxxxxxxxx] Inviato: giovedì 17 giugno 2004 17.59 A: For users of Fedora Core releases; raffaele.ronchi@xxxxxxxxx Oggetto: Re: Kernel 2.6.5-1.358smp Hi, It will help if you tell us what is this program doing. May be in Fedora this program runs faster and ends its task using less time than RH system, of course, at the cost of more CPU cycles ;-) Greetings. El Jueves, 17 de Junio de 2004 17:45, raffaele.ronchi@xxxxxxxxx escribió: > Hello all, > I have installed Fedora Core2 (Kernel 2.6.5-1.358smp) on a system with Xeon > MP processors. I ran the same program (multithreaded) on this system and > compared the results Vs. what obtained on the same hardware running RH > Advanced Server 3.0 (Kernel 2.4.21-9.ELsmp). > > Follows sar output: > > Linux 2.4.21-9.ELsmp (RH system) 06/17/2004 > > 05:00:00 PM CPU %user %nice %system %idle > 05:02:42 PM all 3.82 0.00 1.20 94.98 > 05:02:47 PM all 3.36 0.00 1.45 95.19 > 05:02:52 PM all 3.26 0.00 1.40 95.34 > 05:02:57 PM all 3.01 0.00 1.20 95.79 > 05:03:02 PM all 3.51 0.00 1.60 94.88 > 05:03:07 PM all 3.11 0.00 1.76 95.13 > 05:03:12 PM all 3.30 0.00 1.55 95.15 > > -------------------------------------------------------------------------- - >------------ Linux 2.6.5-1.358smp (Fedora Core2 system) 06/17/2004 > > 04:32:52 PM CPU %user %nice %system %iowait %idle > 04:35:26 PM all 5.05 0.00 6.85 0.20 87.89 > 04:35:31 PM all 5.00 0.00 6.70 0.10 88.21 > 04:35:36 PM all 4.70 0.00 7.10 0.20 88.01 > 04:35:41 PM all 4.90 0.00 6.75 0.20 88.16 > 04:35:46 PM all 5.10 0.00 6.50 0.15 88.24 > 04:35:51 PM all 4.75 0.00 6.95 0.20 88.10 > 04:35:56 PM all 5.74 0.00 8.04 0.20 86.01 > > Both system time and user time are much higher on Fedora. > Is it something you have already seen happening, or might just be related > to kernel configuration ? > > Thanks to everybody will reply. In case I can provide more > details/mesurements. One interesting thing is that the program is running > even slower on Fedora. > > Regards > > Raffaele -- Saludos. --- Carles Xavier Munyoz Baldó carles@xxxxxxxxxxxxxxxxx http://www.unlimitedmail.net/ ---
