Large Upload Speed Difference between Firefox and IE 7

Walter · EVDO Junkie Joined: 14 Sep 2005 Posts: 307

Hi horses. Didn't mean to imply any degradation of your cranial capabilities. Smile

Since I have looked at my TCP/IP settings before and know them to be sound, I pretty much had ruled that area out. You had no way to know this though. So no harm, no foul I hope.

I think I can rule out cache effects as well since the Firefox speeds are repeatable; even back to back. If the cache was having an impact, you would expect the 2nd time around to have everything in the cache already and speed things up. Plus it's unlikely that the writers of all three speedtest sites would make such a basic error.

Since justanoldguy did the exact same test and isn't seeing the same thing as me, then I agree that there must be something about my configuration. I also think this might be a pretty common configuration problem because of the numbers here that have Rev. A, yet report sub 200 kb/sec upload speeds.

Right now I think my best bet is to look through the parameters in the about:config page that rickster mentioned. I'll check that tonight and see what I find. Then I'll probably also do some more FTP tests as well as a cross-check. I'm open to other suggestions that people may have though. And I'll post what I find out tonight. My guess is this may be an issue for a fair number of people.
_________________
Reach, Connect, Celebrate, Grow, Serve.

Michael · Posted: Tue Mar 13, 2007 3:01 pm Post subject:

horsessmellfunny · Posted: Tue Mar 13, 2007 5:37 pm Post subject:

Michael, I am not sure of your Cynicism in your reply, but I certainl do not appreciate it. Many of us are helping people for free here. We dont make any money on our advice. I am only offering knowledge that I have gained over the years of computer science, and over the past year of EVDO. I most certainly do not appreciate your comments.

On a side note, yes it does make sense that the Flash and Java files are not cached (duh on me).
Let me ask this, when you do these speed tests are they flash based or Java based, and in which one are you seeing the most issue. For example, the speedtest.net, (I can remember) is that Flash based? If so, has anybody tried a Java based speedtest and a Flash based test and compared the resuslts side by side from the two?

JoeyDee · Posted: Tue Mar 13, 2007 5:54 pm Post subject:

try http://www.bandwidthplace.com/speedtest/

I'm pretty sure they arn't java or flash Razz

ChrisM · Posted: Tue Mar 13, 2007 7:57 pm Post subject:

I have done extensive work at the corporate level on IP setting optimizations, and applications CAN override TCPIP send and receive window sizes. You need look no further than the settings for the Apache web server.

http://httpd.apache.org/docs/2.0/mod/mpm_common.html#receivebuffersize
http://httpd.apache.org/docs/2.0/mod/mpm_common.html#sendbuffersize

Michael · Posted: Tue Mar 13, 2007 8:22 pm Post subject:

Around 12 years ago, I actually wrote a web server from scratch. Comparing a web server, which of course needs to control many low level settings, isn't really the same as a web browser.

Now, how about we keep this thread on topic.
_________________
EVDO :: EVDO News :: EVDO Antennas :: Buy Verizon :: Buy Sprint :: EVDO Amplifier

Walter · EVDO Junkie Joined: 14 Sep 2005 Posts: 307

Walter · EVDO Junkie Joined: 14 Sep 2005 Posts: 307

Well, I made a first pass through the about:config page in Firefox and so far nothing jumped out at me that might affect upload performance. Might still be there though. I'll keep looking.

Another interesting tidbit of info. I have the IETab addon installed in Firefox. I tried one of my usual speedtest sites. In Firefox mode, still around 200 kb/sec uploads. Switch Firefox to IE mode using IETab, over 500 kb/sec.

Curiouser and curiouser. Confused

_________________
Reach, Connect, Celebrate, Grow, Serve.

Walter · EVDO Junkie Joined: 14 Sep 2005 Posts: 307

Loaded the FasterFox addon. No change in the lower upload speeds in Firefox.
_________________
Reach, Connect, Celebrate, Grow, Serve.

ChrisM · Posted: Tue Mar 13, 2007 9:56 pm Post subject:

Hi, IE7 could be getting better performance than Firefox because it is dynamically changing the TCPIP settings that normally require registry tweaks to enable as the defaults. The idea that applications can't change the IP settings is incorrect. They certainly can. From this article: http://www.microsoft.com/technet/community/columns/cableguy/cg1105.mspx

rickster · EVDO Junkie Joined: 27 Aug 2005 Posts: 216

IMHO You'd more than likely get better advise about this on the mozilla forums I pointed to. The guru's hang out there and I'm guessing any append stating that FF is slower than IE is going to get some quick attention Smile

While this doesn't solve your problems at all I think some other interesting points that may help is try that machine on a faster connection like a wifi coffee shop or a friends DSL/Cable connection.
BTW I've tried here on DSL XP latest IE 7.0, FF 2.0.0.2, and Opera and I see 6M/450 with the normal slight variance between runs. No browser is a winner.

Also un related, the send and receive buffer on an HTTP server are totally un related to any send or receive buffers used by network adapter or IP stack. This is apples and oranges.

ChrisM · Posted: Wed Mar 14, 2007 8:32 am Post subject:

Hi rickster, I'm curious on what basis you make that last statement. AFAIK, an HTTP server running on a desktop uses the exact same network adapter and IP stack as a browser. Why would this be apples to oranges? The point of posting those Apache settings was to demonstrate that an application can modify the IP settings independantly from the OS. This is true for the Apache setting example I gave earlier, as well as for any application running TCPIP on Windows 2000, XP, Solaris or RedHat Enterprise Linux AFAIK.

Operating systems do not know the difference between "server" applications and "client" applications. A program is a program is a program. The user account under which the program runs has far more to do with what the application can do than whether someone has arbitrarily classified the application as a "server" or a "client".

Microsoft, with their complete access to proprietary windows operating system source code could certainly re-write portions of the Windows TCPIP stack specifically for IE7 to optimize browser performance, although this would be "reinventing the wheel", I can certainly see IE7 engineers doing it rather than dealing with trying to get a patch out to every Windows XP desktop. This would further blur the line between application and operating system, something MS has no problems doing, and it would then be possible for them to be able to change settings at the application level that are not even modifiable at the OS level.

Walter · EVDO Junkie Joined: 14 Sep 2005 Posts: 307

Thanks for the tips ChrisM. I think I already made the Tcp1323Opts change a few weeks back when I was doing something else, but I'll check and make sure. One of my regular speedtest sites uses the same style speedtest as the link you provided, but yours has a server closer to me. I think I'll switch to that one as well. I happen to like that particular one as well since it seems to have the least amount of "junk" to load, so you can start the test right away.

Rickster, you're probably right about the Mozilla forums. I quickly checked last night and didn't find any thread like my problem. But a post in those forums with "IE7 Faster than Firefox" in the title *would* get attention I presume. Smile

I'll need to be careful what I say in the text though so I actually get some help rather than just flamed. I may try to post there tonight.

It would be very difficult for me to attach to another fast connection. If I could, I wouldn't be using EVDO in the first place Exclamation

This is a home setup with a server (running Windows XP Pro) and three workstations (all running XP Home). And all three workstations behave the same way.

Personally, I'm beginning to believe IE 7 is turning on some sort of compression. But that's really just speculation. I tried some FTP tests last night to my mail account using a large jpg file (I get a bit of web space as part of the deal so I get a Unix account directory that I can FTP to), but I was getting pretty poor transfers (more like 100 kb/sec) so that path may have had some other limitation. Same with sending an email, although that did look a bit closer to 200 kb/sec. Anyone know of a public FTP server with a good pipe that allows *uploads*? Otherwise, I'll need to arrange with my brother to run an FTP server on one of his machines and temporarily give me access so I can get a decent test off.
_________________
Reach, Connect, Celebrate, Grow, Serve.

Walter · EVDO Junkie Joined: 14 Sep 2005 Posts: 307

Ooooo, I think I just thought of a better test to try tonight. I'm going to upload a good size jpg picture to the Costco photo web site. Once with Firefox and once with IE7. Time it and also watch my NetMeter utility to see what upload speeds are like. That should be a very realistic test and tell me whether the IE 7 upload speeds are *really* higher or not.
_________________
Reach, Connect, Celebrate, Grow, Serve.

rickster · EVDO Junkie Joined: 27 Aug 2005 Posts: 216

If you look at how most web application are written they process requests and form output in very small pieces at a time. Directly doing this to the common tcpip layer and then down to the drivers is extremely inefficient. To avoid this application servers provide on their own caching buffers at the application level.