|
|||||||
![]() |
|
|
LinkBack | Thread Tools | Search this Thread | Rate Thread | Display Modes |
|
|
#1 |
|
Member (14 bit)
Join Date: Mar 1999
Location: Kelowna, B.C., Canada
Posts: 9,138
|
Setting RWIN
Ok, I need to set the RWIN for my freaky cablenet connection. It's NOT where I want it by default, and I have no idea how to change it.
any help? I guess it would help if you knew a bit more ![]() Linux Mandrake 8, Gnome. |
|
|
|
|
|
#2 |
|
Member (13 bit)
Join Date: Jul 2000
Location: Fullerton, CA
Posts: 7,030
|
Hi reboot,
http://www.dslreports.com/tweaks/RWIN I myself use this program to adjust all network settings (RWIN, MTU, etc.) http://www.dslreports.com/front/DRTCP019.exe Hope that helps. Edit: Well, guess the Windows tweaks/software doesn't apply then.
__________________
"A witty saying proves nothing." - Voltaire |
|
|
|
|
|
#3 |
|
Member (14 bit)
Join Date: Mar 1999
Location: Kelowna, B.C., Canada
Posts: 9,138
|
Thanks for trying, but I know how to do what I want in windows, it's Linux this time around
|
|
|
|
|
|
#4 |
|
Premium Member
Join Date: Jun 1999
Posts: 9,231
|
You can use ifconfig MTU size and it works for most chipsets
#/sbin/ifconfig ethX mtu #### |
|
|
|
|
|
#5 |
|
Registered User
Join Date: Mar 1999
Posts: 355
|
RWIN sizes can be tuned, if you have /proc support compiled into your kernel.
the default is 32K, how high do you need it? normally that setting is going to be just fine...in fact, most network settings are going to be optimized for ethernet, unlike windows 98, so there really isn't a lot of tweaking necessary. do you have a high latency connection, or something? anyway, here's a quick run down on the procedure....you can disregard the use of an editor if you want, for the sake of simplicity you could use redirection to change the settings: echo 32767 > rmem_max http://www.psc.edu/networking/perf_tune.html#Linux |
|
|
|
|
|
#6 |
|
Member (14 bit)
Join Date: Mar 1999
Location: Kelowna, B.C., Canada
Posts: 9,138
|
My default seems to be 65535, not 32767, but the TCp/IP analyzer at speedguide.net says it's 5840...who to believe?
|
|
|
|
|
|
#7 |
|
Registered User
Join Date: Mar 1999
Posts: 355
|
I would be inclined to believe:
cat /proc/sys/net/core/rmem_default because, I'm communicating directly with the kernel. if it makes you feel any better, it's detecting my RWIN at 16060, even though the above command is showing 32767. there is one thing to think about, that site seems to be very windows biased, so maybe the analyzer is, too....maybe it's just not up to the Linux TCP/IP stack. drop them a note, and if they don't believe you, I'll drop them a more detailed and annoying note to punctuate the problem. otherwise, tell me what they have to say. you can always try changing the settings and retesting, but remember, the warning they post says that modifying the value may reduce throughput....if the Linux setting is just being misdeteceted, and you tweak the setting up, you could reduce performance in an attempt to improve it. personally, I've sworn off of the tweak of the month club, after having caused myself problems one too many times. |
|
|
|
|
|
#8 |
|
Member (14 bit)
Join Date: Mar 1999
Location: Kelowna, B.C., Canada
Posts: 9,138
|
Well, I am NOT happy with the throughput. It's a high latency connection, and using 65535 in Windows works just fine.
When it was reported at 5840, I had to look. It's at 65535 in ALL the files in /proc/sys/net/core/ but the throughput in Linux sucks. Now I'm thinking it may NOT be anything to do with RWIN, but it might have MUCH to do with Netscrap! and the ad server here at PCMech in particular. Nothing will load until the stupid banner is complete. It's the same way at quite a few other sites. NONE of this happens in Win98 or XP using IE5.5 I've also tried Opera (I HATE adware!!!) and Mozilla as well as Lynx, Konqueror, et al, and they're all based on the same engine, so I get the same results. Just tried konqueror again, and it takes FOREVER to load some websites, with throughput being in the slow 56k modem range (2.5kbps) instead of cable (200kbps). |
|
|
|
|
|
#9 |
|
Registered User
Join Date: Mar 1999
Posts: 355
|
I would think that Netscape might be a good part of the problem. it's the way Netscape/Mozilla handles tables.....the table must load completely before the browser is allowed to move on. the first table seems to make an offsite call to ads1.ad-flow.com, so that's the delay in that ad showing up in the table, and the browser moving on. there's a delay before the banner ad pops up in IE, too, but the rest of the page loads instead of waiting for the banner. not much you can do about it, table's are trouble for netscape/mozilla....nested tables are worse.
there is arachne, which isn't free(read isn't unfettered, but free as in beer for personal use), but it won't be bad when it's done. it's a graphical CLI browser by nature(runs through SVGA-lib), but it also has an X interface. it's also the ugliest thing I've ever seen and doesn't have many advanced features right now. it's still better than Lynx for command line browsing, if nothing else. I'm still working on running IE from my Sun box, it works fine remotely at 8bpp, but chokes when I move the colours up. I need to find a way to fool it into thinking I'm running at 256 colours, I guess. anyway, good luck. |
|
|
|
|
|
#10 |
|
Member (14 bit)
Join Date: Mar 1999
Location: Kelowna, B.C., Canada
Posts: 9,138
|
Thanks, I'm suffering between Konq, Netscrap, and Operadsinyourface for now.
I'll just keep switching back and forth until I settle on one I can live with. For now, it seems MS has the handle on speedy browsing with IE5.5, too bad it just won't live well in my Linux box(es). |
|
|
|
![]() |
| Bookmarks |
| Thread Tools | Search this Thread |
| Display Modes | Rate This Thread |
|
|