SLIP summary posting

From: Randy garrett (rlg@ida.org)
Date: Tue Feb 26 1991 - 16:48:18 CST


Thanks for all the helpful souls who responded. Basically, I wanted to set up a serial IP line over which to run X11 applications. The idea was to determine what sort of bandwidth was required. I was running SunOS 4.1.1 and OpenWindows 2.0 on a Sun 3/60 with 12 Meg of RAM and a Sun 3/50 with 8 Meg of RAM. These are no longer the fastest machines in the world, but they are what I had available for mucking around with their kernels withour undue user complaints. The 3/60 was connected via Ethernet to a number of other Sun 3's and 4's. I acquired slip-4.1-beta from neat.ai.toronto.edu. The instructions are easy enough to follow, so the installation proceeded quickly, but when I tried to use slip-attach to actually connect the 2 machines I got an ioctl error. A quick call to sun-managers resulted in my finding out that there was a typo of one character in slip-attach.c. Well, it was labeled beta software. I am using this version since I was running OS 4.1.1 and I under!
stood the earlier versions of sli

        /* push the SLIP module */
        if (ioctl(sfd, I_PUSH, "slipen") < 0) {
                syslog(LOG_ERR, "ioctl (I_PUSH) slipen: %m");
                exit(1);
        }

My original version had:

        if (ioctl(sfd, I_PUSH, "slipe") < 0) {

instead. Note the 'slipe' vs. 'slipen' . The original site:

dmssyd.syd.dms.csiro.au has this typo corrected.

After this the hookup was simple using slip-attach. I then did rlogins to other machines on the Ethernet using the 3/60 as the gateway o

Basically, I wanted to set up a serial IP line over which to run X11 applications. The idea was to determine what sort of bandwidth was required. I was running SunOS 4.1.1 and OpenWindows 2.0 on a Sun 3/60 with 12 Meg of RAM and a Sun 3/50 with 8 Meg of RAM. These are no longer the fastest machines in the world, but they are what I had available for mucking around with their kernels withour undue user complaints. The 3/60 was connected via Ethernet to a number of other Sun 3's and 4's. I acquired slip-4.1-beta from neat.ai.toronto.edu. The instructions are easy enough to follow, so the installation proceeded quickly, but when I tried to use slip-attach to actually connect the 2 machines I got an ioctl error. A quick call to sun-managers resulted in my finding out that there was a typo of one character in slip-attach.c. Well, it was labeled beta software. I am using this version since I was running OS 4.1.1 and I understood the earlier versions of slip were not direc!
tly compatible. The corrected li

        /* push the SLIP module */
        if (ioctl(sfd, I_PUSH, "slipen") < 0) {
                syslog(LOG_ERR, "ioctl (I_PUSH) slipen: %m");
                exit(1);
        }

My original version had:

        if (ioctl(sfd, I_PUSH, "slipe") < 0) {

instead. Note the 'slipe' vs. 'slipen' . The original site:

dmssyd.syd.dms.csiro.au has this typo corrected.

After this the hookup was simple using slip-attach. I then did rlogins to other machines on the Ethernet using the 3/60 as the gateway or router. The only network connection from the 3/50 was via the RS-232 line to the 3/60.

Helpful commands were netstat -i and netstat -rn. I also configured the slip
machine as a subnet work and used: route add net slip-net ether-net cost,
where slip-net and ether-net were the appropriate IP addresses. I used a cost
of 1.

Anyway, I found the performance of text applications such as vi or emacs using X11 protocols to be quite acceptable. There is a discernable "catch" as a whole page is repainted, but it is very brief.

Graphics is another story. I tried running a few, but a good torture test might be ico which paints a moving, rotating shape in a window. It worked, but slowly. In my case, the limiting factor seems to be the number of interrupts that the 3/50 can field per second. With no load other than OpenWindows and 2 perfmeters, the 3/50 shows less than 50 interrupts per second on the perf meter. Running ico remotely on a Sun4 with the display on the 3/50 shot the interrupts up to around 3000. The 3/60 that was acting as the router shot up to about 6000. Interestingly, the CPU on the 3's was not at 100% on either machine. Also, increasing the baudrate on the RS-232 line from 9600 upto 38.4k did not yield any discernible increase in performance, leading me to believe that the interrupts are indeed the limiting factor. Obviously higher performance CPUs or more intelligent serial ports might alleviate this problem.

I also heard of two other serial line IP programs: ppp and cslip. cslip is supposed to be compressed slip -- presumably it does some type of run-time compression. ppp is supposed to be a newer program altogether. Most respondents felt these two would be superior to slip although there was some sharp dissenting opinions on that. In any case, I have not tried either of these two, although I plan to do so. I recently acquired ppp from: omnigate.clarkson.edu. cslip
is available from: rtsg.ee.lbl.gov .

If these provide interesting results in comparison to slip, I will post a summary.

So many people sent helpful replies that I cannot list them here, although I did try to respond personally to them. Thanks again to the friendly folks on sun-magers.

Randy Garrett
rlg@ida.org



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:11 CDT