Originally Posted by mollda (Post 35265896)

Ok thanks, thought for a minute that VM were giving me extra speed for nothing.

Originally Posted by qasdfdsaq (Post 35266306)

Ooh I like this. Definitely gotta add that to my toolbox, my server host giving me grief with my own UDP test.

Honestly, a UDP test is as far as I'm concerned the only accurate and correct way of doing a speed test, though I've never found a decent automated one before. This is superb. Simply The. Best. Speedtest. Ever.

On another note, VM seem to prioritize UDP traffic (I think) so it's a good measure of how your "last mile" DOCSIS link is performing too.

[Edit]

I cannot stress how amazing this tool is. As long as their capacity remains good, this could well be the best connection speed diagnostic tool in existence.

Originally Posted by kwikbreaks (Post 35265873)

Some AV products can give false results (Kaspersky is/was well known for this and I think the free VM AV is based on Kaspersky)..

Originally Posted by qasdfdsaq (Post 35266306)

Ooh I like this. …..
On another note, VM seem to prioritize UDP traffic (I think) so it's a good measure of how your "last mile" DOCSIS link is performing too.
…..I cannot stress how amazing this tool is. As long as their capacity remains good, this could well be the best connection speed diagnostic tool in existence.

Originally Posted by TJS (Post 35266387)

Agreed, seems very accurate :)

http://mcslhr.visualware.com/myspeed/db/report?id=1153309

Originally Posted by horseman (Post 35266633)

Hmmm - interested to see your evidence to support UDP "prioritisation"….. aside from relative effect of comparing it with protocol shaping on p2p/file sharing that is….. ;)

….and remember like any "test" it still relies on efficacy of VM's peering/transit link choices…..

You might want to experiment with VW's Maidenhead and Dallas servers just to see for yourself perhaps? Here's suitable XXL capacity test links:
Maidenhead >http://mcs.lushsoft.com/myspeed/mycapacity_50mb.html

Dallas > http://mcssjc.visualware.com/myspeed...city_50mb.html

See if you discover any correlation with Dallas & London? ….but then perhaps find that Maidenhead is a little tardy…… or not?

…anyway - have fun!…… ;)

EDIT:


Interesting upstream stats - just out of academic curiosity I don't suppose you're serviced via a Motorola BSR 64k CMTS are you?

Originally Posted by qasdfdsaq (Post 35266679)

3 DS channels. Bingo.

Originally Posted by qasdfdsaq (Post 35266448)

He wasn't taking knocks. It's well known Kaspersky (many versions) affect speedtest results, including making Speedtest.net jump to 200mbps or more on a 20mb line. Speedtest.net have a admitted this and introducted a workaround, many other sites are still affected.

Both the KIS and KAV engines have this effect. Please don't attack others when they post accurate information especially when you don't have a clue.

Originally Posted by TJS (Post 35266691)

Which is the better cmts motorola or cisco? :)

Originally Posted by qasdfdsaq (Post 35266665)

The only evidence I have is doing a HTTP, TLS or SFTP download test with 10-20 streams may only hit 20mbps while sending a single 50mbps UDP stream results in 50mbps data received with minimal loss.

i.e. TCP = ~2mbps per stream, UDP = 50mbps per stream.

Originally Posted by horseman (Post 35266931)

With VM's current configuration :
3 d/s bonded channels = Cisco 10012
4 d/s bonded channels = Motorola BSR 64000

Which is better for what?

BSR 64k is newer kit and can utilise TX32 decoupled downstream modules and RX48 decoupled upstream modules which can gives better flexibility for scalable expansion, smaller physical,power consumption and improved cost/bandwidth profiles.

However it depends on when and how much VM are deploying - you need someone like Igni to give a more authoritative and comprehensive outline of what is out there and how VM are likely to utilise it…..

I only asked whether you might be on Moto kit because of the flat delay line on your u/s capacity that seemed (in my very limited experience) uncharacteristic of Cisco, but obviously the u/s load on your port must presumably be (very) low….. ;)

---------- Post added at 08:31 ---------- Previous post was at 08:24 ----------



….and just to "muddy-the-waters" further a significant difference between single and multistreamed test could also be indicative of congestion! ;)

Originally Posted by qasdfdsaq (Post 35266913)

http://www.lmgtfy.com/?q=Kaspersky+Speedtest

http://blog.ookla.com/2011/05/02/the-kaspersky-effect/

http://forum.kaspersky.com/index.php?showtopic=151143&st=0&p=1216815

http://forum.kaspersky.com/index.php...st=0&p=1283685

So, confirmed by Google, Speedtest.net's makers, and Kaspersky's own moderators. Proof is all over the place if your lazy ass could only be bothered to look for it. I'd quote my own tests too but you seem like the kind of guy who's too ignorant to listen either way.

Originally Posted by Failswitch (Post 35267008)

Once again if you actually bothered to read what you linked to the issue is down to Kaspersky Internet Security and it's network traffic scanning module and has nothing to do with Kaspersky Anti-Virus.