To: E. Davies who wrote (18046 ) 12/19/1999 8:06:00 AM From: matt gray Read Replies (1) | Respond to of 29970
Lets make up an imaginary case. Say the ATHM regional headend is in Portland and the second tier headend is 50 miles away and they have hub points in three other towns. Now an ISP decides to serve these towns and requests a co-location. They order up a DS-3 link to a predetermined meet point in Portland to serve traffic. This meet point is outside the headend in Portland. BTW nothing is really virtual since this virtual link will still require space in the headend in the form of transmission gear to serve the link. The link then interconnects into the ATHM cloud. Since traffic actually flows in multiple locations [in our case a simple small 5 sites] how do we know which traffic comes from which point. All the points have different travel distances and therefore cost different amounts per bit. In certain cases these costs were never formal tarriffs filed with the state but capacity lease agreements since the internet business maybe didn't require tarriffs and other other links are bought from the RBOC and are tarriffed. What does ATHM charge? Futhermore each link from the regional headend to a lower tier headend is part of an ATHM cloud or maybe realistically they are all different services arranged by different carriers some of the links are PPP links others are ATM links. These all have seperate costs associated with them and they were sized based upon certain traffic assumptions. Also the links contain programming info [non ISP related information [eg. Web TV, DTV, analog control channels etc.] which effect link capacity that may not be measurable depending on implementation. Finally we get to the hub location that actually converts the analog bit streams onto the cable system. These schemes [implementations differ between MSOs] are usually contention based. Like the old CSMDA stuff that we all learned in early ethernet theory, as we add packets the overall thruput differs non-linearly because of collisions and retransmission. Furthermore there is no way to monitor who's traffic is causing the link to congest. How does my traffic effect the cost of operating these links? Quality of Service is effected. As ATHM maybe my NOC monitored these links to minimize impact and provision additional circuits or subdivide nodes etc now I have multiple players who have different opinions on what quality of service is not to mention how to fix it. These just touch on a few issues, as the detail is fleshed out many more appear. The whole thing is a mess even in the purest scenario. This is a singular scenario. Throw in some other variables like meet point locations, headends, channelization, server placement, caching, timeframe, etc etc and it gets increasingly complicated. Unlike bell who sells pairs this is a FDM bussed system and each element on the system is effected by each other. Imagine multiple teams of lawyers arguing over each little issue?