Web slowdown with v4.6

Avast Free Antivirus / Premium Security
1

I recently (early last week, I believe) upgraded to the 4.6 beta and today upgraded to the fully released 4.6 version. By and large, I love it! ;D

However (always a catch 'aint there?) I’ve noticed a slight general slowdown on web page loading. I’ve always used Numion’s YourSpeed internet speed measurement page (http://www.numion.com/YourSpeed/Checkup.php?L=world&Duration=30&Repeat=600&Layout=1) to check my surfing speed. It measures true effective browsing speed by downloading 50 website logos from around the world repeatedly for 30 seconds. This provides a world-wide “view” of the internet’s speed as seen by the browser as opposed to D/L a single large file.

I’ve attached a GIF of the last 25 speed measurements. (It’s a graph from Numion’s site.) I use Outpost Pro with the DNS cache enabled so usually the first speed test will be relatively slow since the icon’s URL"s are almost never in the cache. I run the test a second time and, with the URL’s in the cache, the speed typically runs between 350 and 400 Kb/sec. (I’m on a 512/256 DSL line) depending on how busy the internet is at the time.

As you can see, there’s a big dip about half way through the chart which coincides with my installation of the 4.6 beta. As a test, I switched off the proxy settings in IE and re-ran the Numion test and got notably better speeds. (The bump near the right side of the chart) Then I re-enabled the proxy settings and re-tested. Back in the basement. :frowning:

I remember Vlk or Lukor mentioning that Avast can handle way more HTTP requests than a typical home internet connection can process so I’m at a loss to explain the slowdown. It does seem to be consistently slower, however.

Perhaps a few 4.6 users could duplicate my test by switching the proxy settings off and trying Numion’s speed test and post back with their results.

FWIW I run WinME (pauses for laughter to die down) fully updated on a Celeron 1.1 Mhz system with 768 Mb RAM, Outpost Pro 2.5 with DNS cache plug-in enabled.

Strangly enough, the slowdown is hardly noticeable except for Numion’s test site and other such high-activity sites. Normal speed tests using single-file downloads seem just as fast as ever. I have noticed that sometimes not all web page icons load (including this forum at times) since the update to v4.6, but a re-load of the page usually corrects the problem and I’ll not worry about that just yet. :wink:

Any suggestions will be greatly appreciated.

Oh yeah, almost forgot: Thanks Alwil and the great mods around here for the terrific Avast software and the outstanding beta forum. These people rock! 8)

2

I’m beginning to think that I’m the only one with this problem. :cry:

Has anyone tried Numion with/without proxy settings. (I don’t know how you’d do this with WinXP. I guess you’d have to disable WebShield).

Returning to loop.

3

According to the 3 tests I have done on the website you mentioned, my speed is only 1/8 of the maximum.
And I know for sure that ain’t right.

4

Thanks Eddy,

Sounds like what I’m seeing. Actual browsing w/v5.6 feels great! Only occasional slowdowns (not all that unusual during busy web times) but Numion’s test has me puzzled. It’s always been extremely accurate with regard to the “feel” of web browsing in the past, but using WebShield proxy it reports way slow. I know if actual surfing were as slow as the Numion test shows, I’d notice it!

Is it possible that the simple fact of sticking an additional process (proxy) “in the loop” could cause a test such as Numion’s with it multitude of HTTP:GET requests could slow things down?

I think I’ll try leaving IE’s proxy settings pointing to WebShield, but pausing Webshield itself makes any difference. That might help identify if the “slowdown” is cause by the proxy process or by WebShield’s scanning.

I’ll post back in a bit.

5

OK,
I’ve attached a gif of a series of Numion speed tests:

  1. IE using WebShield proxy
  2. IE accessing web directly
  3. IE using WebShield; but WebShield provider paused
  4. IE using re-started WebShield proxy
  5. IE accessing web directly (again)

It certainly appears that simply inserting another process in the HTTP chain is the main cause of the “slowdown.” The slight increase in speed from test (3) to (4) is not significant enough to be more than the usual variation in the web speed. Switching off the proxy in IE jacked the speed right back to the top.

I’ve also attached a firewall (Outpost Pro v2.5) log showing the connection requests from IE to WebShield and then out to the web. I haven’t had much time to study the log myself, but the timings between IE requesting a page and Webshield requesting it from the web server seems disjointed at best. Things actually seem out-of-sequence. It may just be that the system is unable to log the times with sufficient accuracy and is listing things in a different sequence than they actually happened. (The file is a csv using “;” saved as a txt file.)

At any rate. I consider this to be an interesting, academic, question that I’d love to better understand, however I’m not expecting this to be a big priority. Any suggestions will, of course, be greatly appreciated.

Doncha just love new toys? 8)

(I had to drop the last 500 lines of the log to stay below the attachment filesize limit.)

6

no need to post another picture, but: i have similar results to report. webshield seems to absorb 40-50% of native throughput.

7

I just ran this test with Web Shield on and got a result of 370 Kbps. With Web Shield turned off comletely I got 363 Kbps. My results were faster with Web Shield on! :slight_smile: I’m sure that wouldn’t happen all the time if I ran multiple tests but I think it’s clear that I’m not seeing significant slowdowns due to Web Shield on my machine.

2.0 GHz Pentium 4
1.5 Mbps DSL
Win XP Home SP2
Firefox 1.0
Kerio Personal Firewall

8

I just ran several more tests and I was not able to detect a noticable difference in performance in Firefox with Web Shield on vs. Web Shield off.

However, I also ran a bunch of tests in IE. I did notice a minor decrease in IE with Web Shield on, but it was very minor. The most interesting thing was that my speeds were regularly far higher in IE than with Firefox. My average speed in IE was 120 Kbps faster than my average speed with Firefox.

Hmm…food for thought. I always suspected that Firefox was slower than IE.

Now we’re really off-topic! Sorry, Vlk! ;D

9

Its wonderful AVAST, the interaction, response everything.
However, I am also having problems since beta 602. Today I upgraded to the final version.
While everything is great, It is taking a long long time to shutdown and also to browse a few pages which were doing fine till 601.
I Have tried repeatedly the shutdown process which is really taking a lot of time .
I hope the ALWIL team will come out with a solution , as usual, very fast:-[

10

neiby and drahnier,

I suspect that you’re both running WinXP. :wink: I’m running WinMe which is really win98 in sheep’s (hyena’s?) clothing. Avast uses an entirely different method to inject itself into the HTTP data stream for win 2000/XP and I seriously doubt that any Win2000/XP user will see the same impact as us poor win9x users.

As my chart shows, switching off the scanning process (by using the “Pause Provider” option for WebShield) resulted in very little difference in speed. That’s because the proxy chain is still being implemented. IE is sending requests to WebShield which is passing the request through to the web and the returned HTTP stream is passed transparently through WebShield w/o scanning back to IE.

It’s only if I disable the proxy connection in IE connection settings that the speed returns to normal.

This means that the “problem” isn’t really an Avast problem which is why I said that I didn’t expect it to be a big priority for the mods or Alwil. Most folks have/are moving away from Win9x anyway and I sure can’t expect Alwil to fix what appears to be a Win9x problem.

I would, however, be very happy if I, or someone here, could find a way to regain that speed. ;D As I’d noted early in the thread, the actual surfing speed seems only slightly reduced. It’s mostly on Numion’s speed test or other graphic heavy pages that thing’s slow down.

neiby, I suspect your slower Firefox speeds might be due to the fact that (IIRC) Firefox sets itself up as a proxy also. I think its something like: Browser GUI → Firefox proxy → Avast → WWW with the reverse chain on the way back. I think the extra process might account for the slight slowdown.

11

gbark, we have just discussed this with Vlk and we think that the ‘slowdown’ could be possibly caused by those banners that are not working - and I see several of these when I try the speed test from here. WebShield tries to retry all unsuccesfull requests 3 times - similiary as Internet Explorer does - and this may affect the overall speed in this type of measurement.

Could you please try regular throughput test - for example here: http://schema.chello.upc.cz/speed/dwn.php?2000.

12

lukor,

You may be on to something! As I think about it, if IE is retrying 3 times for each GET that times out, and WebShield is doing the same thing, then each unsuccessful initial GET from IE would generate 9 GETs from WebShield! It sure seems like that ought to slow something down.

I’m not sure just how that would affect Numion’s speed calculations, since the javascript used would only be aware of the GETs made directly through IE and the timer would be counting just those requests. Interesting interaction, to be sure. I think I’ll post a link to this thread to Numion’s tech support. They may be able to shed some light on the situation. Numion’s FAQ and “Technical Tidbits” page has lots of good information on the process their speed test uses.

I tried your speed test link. It reports upwards of 410Mb/sec for my 512 Mb/sec DSL connection. Considering DSL and other overhead, that’s about normal. Actually Numion’s “MaxSpeed” test and DSLReports speed tests and most other single, large file download type tests typically report between 410 and 460Mb/sec. This has changed very little since my upgrade to Avast 4.6. It’s obvious that WebShield (and the proxy process in Win9x) has little impact on this type of activity.

Why does WebShield perform 3 requests on failed connections if the browser is also doing it? Can that be changed through the Avast4.ini file?

Oh well, I love a challenge! Thanks for the input.

And thanks again for a great piece of software. 8)

13

Gbark, when internet explorer connects to webshield, for it’s side the connection has already succeeded. Now, the other end connection (between webshield and the real server) is going to be established. If this connection fails, it’s the webshield’s task to retry that if possible. Internet Explorer retries a little bit more aggressively that e.g. Firefox. So there are indeed situations where 9 retries take place, instead of 3, but we thought that it is already the bad case when the link is dead and shouldn’t bother the user during regular browsing.

Currently there is no option in the avast4.ini to configure this. Should we add some?

Lukas.

14
Currently there is no option in the avast4.ini to configure this. Should we add some?
No, since this "slowdown" probably only affects win9x users, and (at least for me) doesn't impair normal browsing, just certain speed tests, it seems unreasonable to make changes. Besides, I can always simply turn off the proxy settings in IE and run the test. No need to reboot or even exit IE for this. When the test is complete I simply re-enable the proxy settings. :)

I haven’t heard from the Numion folks since I posted to their support page. If I receive any interesting information from them I’ll post it here.

If it were affecting normal browsing, or if more users reported related problems, we should work on a solution, but I don’t think it warrants spending any more of your valuable time. I’ll keep this thread tagged for reply notifications, however.