Friday, 15 August 2014

What Does a 5 Mbps Target Really Mean

Connecting Canadians is one of the five pillars of the federal government's Digital Canada 150 plan announced on 22 Jul 2014. 

A main objective within the Connecting Canadian pillar is to ensure 98% of all Canadians will have access to high-speed Internet at 5 Mbps. As presented in the government paper the blog entry the day of the announcement, this is a rather meaningless statement as no measurement standard for the 5 Mbps is used

In the blog entry of 29 Jan 2012 entitled What Mobile Speeds Really Mean I discussed the three different ways mobile data speed can be measured: - Theoretical Maximum, Vendors Advertised Speed, and Actual Speeds Available to the User.  If I were writing the article today, I might be inclined to add a fourth category, namely, Speed Identified in a Government Funding Program.  

While the original item was directed at Mobile (Cellular) systems, the same three classifications can be applied to terrestrial (cable and DSL), fixed wireless and satellite.

The problem with public Internet connectivity networks is that most of them operate as a shared networks as some point.[1] Based on a net neutrality model, this means that all the active users are sharing all the available bandwidth simultaneously.[2] Assuming a fixed bandwidth model, say 5 Mbps, this means that as the number of active users increase, the amount of bandwidth available to service each active user decreases. Eventually the network reaches the point where it collapses under its own weight.  

This phenomenon is most likely to occur during the peak usage period, some times referred to as the rush hour. While the busy period can vary from location to location, the industry generally accepts the period of weeknights between 7:00 pm to 11:00 pm local time.[3]

Should we allow an ISP to state a 5 Mbps speed if it cannot deliver anywhere near that speed during the peak usage period? I say no.

The US Federal Communication Commission (FCC) has been tracking the real speed provided by US Internet Service Providers (ISP). They have come up with the terms "Advertised Speeds' as used by the ISPs to market their product and "Sustained Speed" which is designed to describe a long term average for broadband (high-speed Internet). The FCC study concludes that speeds can only be stated as speeds obtained during the peak usage period.

Researchers at North Carolina State University (NCSU) developed a method of measurement that allows for a base metric. The NCSU system is based on a metric designed to convey how likely any given consumer is to experience broadband speeds of a particular level. The key to the system is the percentage of users receiving a percentage of the advertised speed. This results in numerical score.

The first step is to identify a target level based on the formula:  percentage of users for percentage of the time during peak hours.  Targets levels are expressed as 80/80 [4] where 80 percent of the users get 80 percent of the advertised speed in a fixed period. FCC has been using the 80/80 target level.

This information is monitored by special routers install at the user location and reported to the FCC. An example of final calculation might be 78% of user met the threshold target of 90/90.  This ISP would be rated as 78%. The ISP could raise the score by increasing the bandwidth available during peak hours. However, it would still only be able to advertise a speed of 78% of 5 Mbps until they met or exceeded 100%.

As part o the FCC study, there were ISPs delivering service in excess of 100%, i.e. they consistently delivered speeds in excess of those contracted. The highest rated was the satellite service Exede while the most of the others were fibre optic based.

The Digital 150 plan needs to adopt an approach similar to the FCC. Allowing an ISP to say they are delivering 5 Mbps while actually providing about 20% of that speed to the majority of users is not good enough.

A good start would be the use of the term "sustained speed" as the 5 Mbps target speed.

[1] While most people have experienced cellular site congestion at the local level, recent studied have indicated that major congestion also occurs at Tier 1 peering points which would affect all type of networks. 
[2] There are network controls available that can adjust access and assign priority to selected users. Some, such as Quality of Service (QoS) that assign priority to real-time applications like VoIP are acceptable. Others, such as "throttling" violate net neutrality concepts or are frowned upon.

[3] There is also anecdotal evidence in the Algoma District that the peak usage period closely follows the hours students are out of school including weekends.

[4] Any threshold ratio can be used such as 70/70 or 90/90. The lower the ratio, the more likely a higher score will be achieved.