|
|
|
Re: slow tcp on 100mbps wan, strange tcp window behaviour on ubuntu 9.10 | |
| [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] |
|
On Thu, 2010-03-11 at 00:59 +0100, Andres Meyer wrote: > On Wed, 2010-03-10 at 15:49 -0800, Stephen Hemminger wrote: > > If you use tcptrace to produce xplot graphs of sequence number, > > you get the attached result. The upper line is the sequence number > > of the upper bound of the window. > > > > It shows the window for the sender is open but is not being used. > > This means the send window is not configured large enough on the sender. > > With current Linux kernels, you should not have to configure socket > > send buffers in the application. It maybe that the application is > > using setsockopt to set send buffer space but not allowing for enough window. > > Thank you for the tip. I also studied this graph (and others) before > posting. I thought the same thing in the beginning, so I tried setting > the settings in the /proc/sys/net/ipv4 region in the kernel. I tried > with iperf and netcat. I even set the window with the -w option to > 1Mbyte in iperf (client and server), just to see if something would > change, but nothing did. Do you know another program that for sure does > not use this setsockopt for testing? I also tried with wget (client) and apache2.2 (server), same effect. -- To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
[Netdev] [Ethernet Bridging] [Linux 802.1Q VLAN] [Linux Wireless] [Kernel Newbies] [Security] [Linux for Hams] [Netfilter] [Git] [Bugtraq] [Photo] [Yosemite] [Yosemite News and Information] [MIPS Linux] [ARM Linux] [Linux RAID] [Linux PCI] [Linux Admin] [Samba] [Video 4 Linux] [Linux Resources]