When a study simultaneously considers requests with heterogeneous bitrate granularity, the capacity fairness problem arises. Then, a unique blocking probability for all requests becomes unfairness, and it needs to be sub-divided into n-classes.
So, my question is:
Can Bandwidth Blocking Ratio (BBR) overcome this unfairness problem?
In other words, is BBR a fairness metric for requests with heterogeneous bitrate granularity?
This links will be helpfull
https://books.google.co.in/books?id=Vugrgn8jtBgC&pg=PA92&lpg=PA92&dq=relationship+between+Bandwidth+Blocking+Ratio+and+capacity+fairness&source=bl&ots=s8v08nCO3O&sig=Yb1bYCjNHRVjvKME0il4nji1Xzc&hl=en&sa=X&ved=0ahUKEwjw59uYwb7RAhWBQI8KHUFCDIgQ6AEIJTAB#v=onepage&q=relationship%20between%20Bandwidth%20Blocking%20Ratio%20and%20capacity%20fairness&f=false
http://dlia.ir/Scientific/e_book/Technology/Electrical_Nuclear_Engine_Electronics/TK_5101_6720_Telecommunication_/022272.pdf
Dear Kelvin,
A heterogeneous network is said to be fair whether all types of services perceive a similar blocking rate (i.e. intentionally blocking requests from some services to accomodate another services). The most classical metric for fairness is Jain's index (see attached publication), where x_i is related to the blocking rate of service i (adapting to your problem's terminology). Values of Jain's index closer to 1 indicate more fairness.
The blocking probability (I am using rate and probability interchangeably) of each service (BP_i) is simply the quotient between the number of allocated connections over the whole set of requests. So, you can use these values to compute Jain's index considering x_i = 1 - BP_i.
Then, you can compute the overall network blocking probability as "sum(BW_i * BP_i) / sum(BW_i)", where BW_i is the bitrate of service i.
Anyway, your definition of fairness may vary according to your scenario, so you can use this information as baseline.
Best regards,
José Luis
Article A Quantitative Measure Of Fairness And Discrimination For Re...
- Badges
- Science topic