200902122343.jpgI’ve only just now had a chance to start to summarize my thoughts on documents related to CRTC Public Notice (PN) 2008-19; Review of the Internet traffic management practices of Internet service providers that have been filed since January 26th, 2009. Below are points of interest that come up – my hope is in the next few days to integrate and update the initial summary document that I prepared for ISP filings, so that a more complete picture of what has been filed exists.

January 26, 2009 ISP Filings

These filings, by major Canadian ISPs, were in response to the earlier inquiries made by non-ISP interrogatories for the public notice. I put together a summary document concerning those inquiries, and wrote a post that pulled together interesting comments that emerged from them.

Cogeco noted hat it was well known that there was a growth in Internet data traffic, though was not willing to disclose their actual growth numbers. Bell and MTS Allstream both supported the suggestion that the CRTC aggregate raw data traffic information that was provided by ISPs, so long as the information was anonymized and thus kept trade secrets relatively secret. Bell suggested that such aggregations could be divided according to ‘HTTP/streaming’, ‘P2P’, ‘UDP’, and ‘Other’ categories. MTS Allstream suggested that aggregated numbers be divided by ‘Telcos’ and ‘Cable providers’, or by ‘ISPs that throttle’ and ‘ISPs that don’t throttle traffic’.

Bell did not want to release the name of their DPI vendor, or the keys that they use for filtering purposes, stating that this would amount to revealing confidential and sensitive business practices.

Bragg was willing to release a percentage of total traffic in the top 5% and 10% user categories, but was not willing to reduce actual (empirical) amounts of data traffic that these groups consume. As was the case in most of the filings, Bragg insisted that releasing empirical data points would threaten their ability to compete in the market, and run the risk of creating misinterpretations of this data among its current, and potential, customers.

Sasktel released a percentage break-down of traffic-types, which revealed that HTTP/TCP 80 amounted to 45% of their traffic. P2P appears to have accounted for less that 10% of traffic, though given that Sasktel was providing data based on ports, and many P2P programs randomize ports, it stands to reason that P2P traffic actually accounted for a larger amount of traffic than suggested.

February 4, 2009 CRTC Response

In light of the requests for comments by interested parties, and the responses by ISPs on January 26, 2009, the CRTC laid out what information ISPs would have to provide to subsequently be put on the public record. Much of this data would be put on the record in an anonymized format, so that the public could develop informed opinions and comments, while not jeopardizing ISPs’ ability to compete in the market. In the next few days, I’ll be collating the CRTCs requests alongside the public data, so that it is a bit easier to associate ISPs with anonymous numbers.

February 9, 2009 ISP Response to CRTC

I’m going to just quickly sketch out some of the more interesting points that came from these filings.

  • Bell: They define their peak period roughly as “the behaviour of a massive user base using the network at the same time”. Additionally, they imply that wholesaler customers are consuming a disproportionate amount of bandwidth, and have actually doubled their consumption from the point prior to Bell having deployed traffic shaping. Bell has offered three reasons for why they think this has happened. The first was filed in confidence, the second is associated with the use of HTTP for content sharing, and the third based on the typical behaviour of the base of users to whom wholesalers market to.
  • Sasktel: They have identified their peak according to the hourly volume that is calculated on a daily basis. They have noticed that the number of peak periods have increased in the past five years.
  • Bragg: Their peak times are between 6:00pm – 11:00pm, and have remained consistent since 2006. They explicitly note the following applications as “high bandwidth applications” that are latency insensitive: Bit Torrent, News, DirectConnect, P2P (Blubster, Gnutella, KaZaA, WinMX, eDonkey, Filetopia, Hotline, GuruGuru, Soribada, Soulseek, Ares, JoltID, eMule, Waste, Konspire2b, ExoSee, FurtherNet, MUTE, GNUnet, Nodezilla).
  • Shaw: Their peak periods of use are 16:00-2:00, and this has not changed since 2006. While they do not reveal their growth numbers to the public record, they note that their growth is generally in alignment with Cisco’s IP traffic forecast 2006-2012. As a consequence, we can reasonably assume that Shaw was moving around 4, 234 PetaBytes of data in 2006, and expects to be moving around 43,551 PetaBytes in 2012. Shaw has revealed that they use Arbor-Ellacoya equipment, but this technology is not used to examine anything beyond the packet header. When the equipment identifies P2P traffic, it slows it by XX% based on the aggregate use per user. As such, they do not block P2P, though they do throttle upload traffic (not download traffic). Their system affects both retail and wholesale customers, and has ultimately led their upstream traffic to being at a more manageable level than prior to using the equipment.

February 11, 2009 Anonymized ISP data from CRTC

Data that was released by the CRTC is available on their website. I’ll be digging into the numbers more deeply shortly – Paul Jay over at the CBC has a piece that begins to get into the numbers.