M-Labs data and Cogent DSCP markings

[Kilmer, Hank]

Due to the severe level of congestion, the lack of movement in negotiating possible remedies and the extreme level of impact to small enterprise customers (retail customers), Cogent implemented a QoS structure that impacts interconnections during the time they are congested in February and March of 2014. Consistent with recommendations from BITAG (Broadband Internet Technical Advisory Group: http://www.bitag.org/documents/BITAG_-_Congestion_Management_Report.pdf), Cogent prioritized based on user type putting its retail customers in one group and wholesale in another. Retail customers were favored because they tend to use applications, such as VoIP, that are most sensitive to congestion. M-Labs is set up in Cogent’s system as a retail customer and their traffic was marked and handled exactly the same as all other retail customers. Additionally, all wholesale customers traffic was marked and handled the same way as other wholesale customers. This was a last resort effort to help manage the congestion and its impact to our customers.

Hank Kilmer
Cogent

[Livingood, Jason]

Thanks for shedding light on this, Hank! Some replies inline below - this is very interesting.

- Jason

On 11/5/14, 10:51 AM, "Kilmer, Hank" <HKi...@Cogentco.com> wrote:

Due to the severe level of congestion, the lack of movement in negotiating possible remedies and the extreme level of impact to small enterprise customers (retail customers), Cogent implemented a QoS structure that impacts interconnections during the time they are congested in February and March of 2014.  Consistent with recommendations from BITAG (Broadband Internet Technical Advisory Group: http://www.bitag.org/documents/BITAG_-_Congestion_Management_Report.pdf),

Since you mention the BITAG recommendations:

1 - This may be an item worthy of a “review request” at the BITAG. Would Cogent be willing to submit these practices or this issue for review & discussion at the BITAG (which is just wrapping up a paper on interconnection practices in fact)? Could be timely and interesting.

 

2 - Recommendation 7.1 on page 43 is on transparency / disclosure. Was this disclosed by Cogent at that time or is this the first time? Can you disclose all the details? This seems probably the single most important recommendation of that BITAG report.

3 - Recommendation 7.3 says essentially to be careful if network management practices that could detrimentally impact certain users or applications. Are there implications for this one?

Cogent prioritized based on user type putting its retail customers in one group and wholesale in another.  Retail customers were favored because they tend to use applications, such as VoIP, that are most sensitive to congestion.  

That’s interesting. So to use an example, if a retail customer collected a packet capture of different types of traffic they would not see 0x00, 0x28, or 0x48, just the highest 0x48? 

Also, would you have classed Netflix traffic (which I assume is considered wholesale) below the retail traffic (in which the M-Labs traffic was classified)? Such as in 0x00 or 0x28? 

Thanks!

Jason

M-Labs is set up in Cogent’s system as a retail customer and their traffic was marked and handled exactly the same as all other retail customers.  Additionally, all wholesale customers traffic was marked and handled the same way as other wholesale customers.  This was a last resort effort to help manage the congestion and its impact to our customers.

Hank Kilmer

Cogent

--

You received this message because you are subscribed to the Google Groups "discuss" group.

To unsubscribe from this group and stop receiving emails from it, send an email to discuss+u...@measurementlab.net.

To post to this group, send email to dis...@measurementlab.net.

Visit this group at http://groups.google.com/a/measurementlab.net/group/discuss/.

[Paul Wall]

Don't do it, Hank, it's a trap!

Drive Slow,
Paul WALL

[Livingood, Jason]

Paul – A trap would imply that Hank had something to hide, which isn't the case. In fact I think it’s a great first step that he’s confirmed the M-Lab finding of Cogent’s prioritization change. That he’s open to discussing it here (or anywhere for that matter) can only be a positive IMHO. 

Obviously researchers are going to have to digest this a bit once they have more information and work to better understand it, confirm their understanding via relevant PCAPs and so on. A good next step is fully explaining how it works. 

Based on my own experience, born of helping dig Comcast out of a Bit Torrent hole (ug!), my recommendation is to over-index on disclosure. For example, here is how I helped write-up our network management practices in 2008 - http://tools.ietf.org/html/rfc6057, and then we put it all on a site all about network management practices at http://networkmanagement.comcast.net/.

I know the IETF meeting is next week (nerds in paradise 2.0!). Any Cogent folks going? If so, maybe a little bar BoF with folks working directly or indirectly with M-Lab would be interesting? 

Jason

[Matt Tooley]

As one of the contributors to the BITAG report on congestion management techniques, one of the key recommendations in the group's report is for network operators to be transparent about its network and congestion management techniques.  So it's great that Cogent is now sharing this with everyone as it now makes easier for researchers, network engineers, and consumers to better understand what is going so that they can act accordingly.

And and bar BOF at IETF next week would be great.

-- Matt

[Livingood, Jason]

The bar BoF at next week’s IETF is still an open offer. If folks are interested, indicate your so here:

http://doodle.com/472knhcpm7yisui4 

Thanks!

Jason