TheMG
join:2007-09-04
Edmonton, AB
 ·TELUS
1 edit | reply to Name96
Re: How much Bell's throttling affects our network and others
said by Name96  :What's all that undifferentiated UDP traffic, anyway? My guess would be gaming that is not being detected as gaming, VoIP, and other streaming/real-time media. |
|
TSI Gabe
Premium,VIP
join:2007-01-03
Chatham, ON
| said by TheMG :said by Name96 :What's all that undifferentiated UDP traffic, anyway? My guess would be gaming that is not being detected as gaming, VoIP, and other streaming/real-time media. DNS uses UDP
--
TSI Gabe - TekSavvy Solutions Inc. |
|
BozoTheCl0wn
join:2005-04-01
H0H0H0
| said by TSI Gabe :DNS uses UDP Yes but your graph has a separate color for DNS...
The three graphs obviously don't represent the same portion of your network. The first one is in the Gbps while the other two barely go over 100Mbps. If Bell only throttles P2P then your two colored graphs make no sense especially considering several reports of people experiencing full HTTP/NTTP/etc. speeds while P2P being stuck in the low 10's of kbps.
BTW, your "blip" is about 450-500Mbps, nowhere near 1Gpbs as you stated (or did you add in+out?). |
|
TSI Gabe
Premium,VIP
join:2007-01-03
Chatham, ON | It's a netflow graph. Only a certain percentage of the data is collected. But it still represents a global view of the network.
--
TSI Gabe - TekSavvy Solutions Inc. |
|
TSI Gabe
Premium,VIP
join:2007-01-03
Chatham, ON | We should be peaking at 3gbps. We aren't anywhere close to it now.
--
TSI Gabe - TekSavvy Solutions Inc. |
|
DarkStar33
join:2008-03-27
Toronto, ON
·TekSavvy Solutions..
1 edit | reply to BozoTheCl0wn
The graph "looks" the same but take a look at the numbers on the side. Thats the important part.
Last weeks UDP traffic alone is pretty close to the total amount of traffic flowing through the pipes now.
I feel bad for your call center people because I am sure they are fielding complains. Bring them a coffee or buy them lunch, I remember what it was like to be in there position.
Perhaps a breakdown of how TSI classifies traffic by protocol would be helpful to foster understanding. |
|
BozoTheCl0wn
join:2005-04-01
H0H0H0
| reply to TSI Gabe
said by TSI Gabe :It's a netflow graph. Only a certain percentage of the data is collected. But it still represents a global view of the network. Sampled Netflow is pretty common but your graphing applications should be adjusted according to the sample rate you are using if you want the traffic levels to be more representative of reality. Also, Netflow is pretty lousy at identifying applications because it only knows about ports. I wouldn't trust it's P2P classification because BT/P2P apps tend to use ports all over the place... That's why they rely on DPI now because it goes further than just ports and looks at payload signatures to identify applications. |
|
the cerberus
join:2007-10-16
Richmond Hill, ON
| said by BozoTheCl0wn :I wouldn't trust it's P2P classification because BT/P2P apps tend to use ports all over the place... That's why they rely on DPI now because it goes further than just ports and looks at payload signatures to identify applications. Gabe, thats what I was about to ask, considering users that are maxing their torrents, most of them are using private trackers that block common p2p ports such as 6881-6889. I know If I'm helping someone set up port forwarding I never use these ports simply because some trackers block them. So, is this true? does it only look at port numbers? If so i'd say this graph is trash. |
|
Snickerdo
Premium
join:2001-02-28
Niagara Falls, ON
| said by the cerberus :Gabe, thats what I was about to ask, considering users that are maxing their torrents, most of them are using private trackers that block common p2p ports such as 6881-6889. I know If I'm helping someone set up port forwarding I never use these ports simply because some trackers block them. So, is this true? does it only look at port numbers? If so i'd say this graph is trash. Even if you combine all UDP traffic and all P2P traffic on those graphs, you're still at 50%, not 70-90% like Bell has tried to claim in the past.
--
I swear that I will faithfully and bear true allegiance to Her Majesty Queen Elizabeth the Second, Queen of Canada, Her Heirs and Successors, and that I will faithfully observe the laws of Canada and fulfil my duties as a Canadian citizen. |
|
the cerberus
join:2007-10-16
Richmond Hill, ON
| said by Snickerdo :Even if you combine all UDP traffic and all P2P traffic on those graphs, you're still at 50%, not 70-90% like Bell has tried to claim in the past. I'm not so sure about that, I'd like to see a difference (before - after) graph if possible, so we can see exactly whats changed. Any chance we could get some raw charted data (perhaps using a better system then port identification), instead of a graph? then we could make our pie graphs and such for clarity. |
|
LiQuiD
BSD geek
Premium
join:2002-08-08
Anjou, QC
| reply to BozoTheCl0wn
said by BozoTheCl0wn :said by TSI Gabe :It's a netflow graph. Only a certain percentage of the data is collected. But it still represents a global view of the network. Sampled Netflow is pretty common but your graphing applications should be adjusted according to the sample rate you are using if you want the traffic levels to be more representative of reality. Also, Netflow is pretty lousy at identifying applications because it only knows about ports. I wouldn't trust it's P2P classification because BT/P2P apps tend to use ports all over the place... That's why they rely on DPI now because it goes further than just ports and looks at payload signatures to identify applications. Good information to know. That would explain why there is that high amount of "UDP" showing on the graph, when streaming media has it's own category. Worth noting though is that even if you were to combine UDP and P2P together, that still isn't the "80-90%" Bell claimed being used by P2P |
|
Name96
join:2008-03-28
| reply to Snickerdo
said by Snickerdo :Even if you combine all UDP traffic and all P2P traffic on those graphs, you're still at 50%, not 70-90% like Bell has tried to claim in the past. Torrent data transfers use TCP, not UDP. A traffic analyzer that couldn't tell the difference between TCP and UDP packets would best be described as severely broken.
The only things that make heavy use of UDP are streaming media, VOIP (VOIP is really just a subtype of streaming media anyway), gaming and certain VPNs. Nothing else should come close to this level of UDP usage. I'm certainly not aware of any P2P apps that make extensive use of UDP for data transfer. |
|
Snickerdo
Premium
join:2001-02-28
Niagara Falls, ON
| said by Name96 :Torrent data transfers use TCP, not UDP. Not all torrent data is TCP, there is a good chunk that is also UDP, albeit your point still stands. |
|
LiQuiD
BSD geek
Premium
join:2002-08-08
Anjou, QC | So then all that UDP has to be the pr0n streams!! Lord knows the net is consumed by porn. |
|
