Kitz Forum
Broadband Related => ISPs => Topic started by: Weaver on July 11, 2018, 06:31:47 AM
-
Regarding AA’s internal congestion report table (https://clueless.aa.net.uk/linkreport.cgi), I am just being thick here - I don't understand the numbers in the 24-hour and non-premium columns. Presumably 24-hour means all lines, not just non-premium ones, so the figure in the 24-hour coulumn should be some error_count divided by a larger n_premium_lines ?
-
interestingly aaisp rate limit non premium lines if there is congestion, I didnt know that but I am fine with it. I guess similar to enta's old ALT system or whatever it was called but in smaller increments.
The unerrored seconds is if they drop a packet if they do it counts as an errored second, so isnt related to DSL errors. Its across their network.
-
Can you provide a link showing where AA rate limit non-premium customers please? I'm aware that, in common with BT Wholesale, they preserve premium marked traffic over non-premium but this isn't rate limiting.
-
Isn't most of what they do optimising the use of individual lines? If the line is being saturated by huge downloads things like VoIP need to be prioritised.
-
What I don't understand is the relationship between those two columns.
-
Non premium is when packets are dropped due to congestion on non premium lines, Errored seconds is when packets are dropped on their network.
-
Can you provide a link showing where AA rate limit non-premium customers please? I'm aware that, in common with BT Wholesale, they preserve premium marked traffic over non-premium but this isn't rate limiting.
from the link weaver provided
We also rate limit non-premium lines in gradual increments to manage the capacity.
-
Thank you. Interesting.
-
But I don't understand how those fractions can be correct if the group of users is smaller? Do you see the problem re my first post? Tell me if I am just being mad.
-
They don't divide error count by users, this is not a user facing metric, it's on interconnects so user count irrelevant?
That's the proportion of the respective time periods when congestion was seen.
-
Sorry, I should not have said users in my first post, I should have said lines. I am not understanding. They have to count which lines are premium and which are not and since the number of premium lines is less than the total number of lines then that is why I cannot make the fractions fit for the last 14 days row. Does anyone else see what I am seeing? remember that I am full to the eyeballs with drugs so please bear with me if I have got everything backwards, but on one row the inequality is one way around, while on another row it is the opposite way around, yet the one data set, premium, is a subset of the other, the total.
-
If it is any concellation I dont understand the chart either. :lol: One thing I did take away from a quick glance is that TTB looks a more reliable network than BT.
-
Thank you banger, I thought I was going mad. still could well be anyway, regardless.
-
If it is any concellation I dont understand the chart either. :lol: One thing I did take away from a quick glance is that TTB looks a more reliable network than BT.
I read it as A&A need to increase their BT backhaul rather than whether one network is more reliable than another.
-
I took that link (https://clueless.aa.net.uk/linkreport.cgi) and tried to see if I could understand what it shows. Unfortunately I can not understand it. ???
I make the observation that it originates from A&A (https://aa.net.uk/) and, therefore, must be right. (A Felis catus (https://en.wikipedia.org/wiki/Cat), such as I, has no right to understand the workings of such superior minds.)
-
I think they must be using aggregated data for the figures rather than the raw data, and so this can happen https://www.theregister.co.uk/2014/05/28/theorums_3_simpson/ although I can't see how :D. The first one is the one that's really wrong as you can't have a lower "average" number over 24 hours when the congestion only affects part of the group (the Non-Premium).
24 hour un-errored seconds Non-premium Premium 24 hour - non premium
99.62% 99.69% 100.00%
seconds 602501.76 602925.12 604800 423.36
99.74% 99.73% 100.00%
seconds 1206455.04 1206334.08 1209600 -120.96
99.68% 99.64% 100.00%
seconds 2411458.56 2410490.88 2419200 -967.68
-
Boozy - you are a star. Perhaps some of the numbers are in the wrong columns, transposed?
-
It makes perfect sense to me.
You need to explain what doesn't make sense to you?
There isn't and shouldn't be a perfectly logical pattern between the 24 hour un-errored seconds tab and the non-premium tab. They are measuring 2 different things.
A single 100 second sample (more than 0.1% of the day) on the non-premium tab could be made up of 5 errored seconds or 100 errored seconds.
So the % on the unerrored seconds tab can be higher or lower than the % on the non-premium tab.
Or am I misunderstanding your confusion? I'm confused at your confusion :lol:
-
Yeah I can understand it, I dont understand the confusion.
-
I'll try and do this using ice creams, so as not to cause confusion about definitions ;D
you have X people and not enough ice creams to go round, then the chance of getting an ice cream <100% (call it z%).
If you take a bunch of the "premium" people and give them all an ice cream, for the remainder the chance of getting an ice cream is now worse (call it y%). y<z.
i.e. when you remove a portion of the people from a ratio calculated using them (and say they were not affected by the rationing) the ratio can only change in one direction. In the table the change goes in both directions.
The assumption is that the same thing is being measured across the whole table.
-
It is the both directions thing that I cannot understand.
I am assuming that one thing is a subset and the other thing is the entire set, then n(subset) > n(whole set) is impossible.
If there were no premium errored sec events then how can that be added to the non-premium count to produce a total that is higher? as shown in one row at the top
-
I meant, what exactly are the figures you have a problem with.
if it's the 24 hour un-errored seconds and Non-premium tabs (I think that's what you's mean?) then it's entirely possible for 1 to be higher or lower than the other.
They are measuring 2 completely different things.