2.2 to 2.4... serious TCP send slowdowns

From: Hayden Myers (hayden@spinbox.com)
Date: Fri Jul 19 2002 - 12:33:28 EST

 We're finally migrating to the 2.4 kernel due to hardware
 incompatibilities with the 2.2. The 2.2 has worked better for us in the
 past as far as our application performs. Our application is an adserver
 and becomes bogged down in 2.4 when sending files such as images across
 the wire. They're in general between 20-50k in size. I've been
 researching the differences between 2.4 and 2.2 and have noticed that a
 lot of work has gone into autotuning with 2.4 and I'm wondering if this is
 what's slowing things down. When I do tcpdumps to see the traffic being
 sent to the client I'm noticing that the receiver window is almost always
 set to 6430 bytes. When looking at the same transfer on our 2.2 boxes the
 receiver window is almost always over 31000 bytes. I've tried to increase
 the size of the buffers using the proc settings that are provided however
 this hasn't seemed to make a difference even after restarting servers
 after each change the window is still 6430 bytes. I've tried manually
 settting the size with setsockopt calls in the server code but this hasn't
 seemed to help. I believe the problem is definately with sending the
 files over the line. We files are read into the socket to be sent across
 the network byte by byte. The boss says this is the best way to do it but
 I'm curious if this is so. The code that reads the file into the socket
 to go across the network is below.
 int output_block(int socket, char *filename)
 int fd, count = 0;
 size_t total_bytes = 0;
 /*size_t buf_cnt = 1460;*/
 size_t buf_cnt = 512;
 char buffer[buf_cnt];
 fd_set rfds;
 struct timeval tv;
    if ((fd = open(filename, O_RDONLY)) < 0) {
       //fprintf(stderr, "Unable to open filename: %s\n", filename);
    while ((count = read(fd, &buffer, buf_cnt)) > 0) {
       FD_SET(socket, &rfds);
       tv.tv_sec = 10;
       tv.tv_usec = 0;
       if (select(socket+1, NULL, &rfds, NULL, &tv) <= 0) {
          //fprintf(stderr, "Output_block timeout\n");
       if (writen(socket, buffer, count) <= 0)
       total_bytes += count;
 The application is a single threaded app using a multiprocess pre forking
 model if that helps any. I'm really baffled as to why using the 2.4
 kernel is slowing us down. Any help is appreciated. Sorry if this has
 come up before. I really have been looking for help for quite some time
  before posting this.
 Hayden Myers
 Support Manager
 Skyline Network Technologies
 (410)583-1337 option 2

To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

This archive was generated by hypermail 2b29 : Tue Jul 23 2002 - 22:00:01 EST