24.11.2012 Views

Traffic Management for the Available Bit Rate (ABR) Service in ...

Traffic Management for the Available Bit Rate (ABR) Service in ...

Traffic Management for the Available Bit Rate (ABR) Service in ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Observe that this de nition means that <strong>the</strong> optimal vector is such that its smallest<br />

component is maximized over all feasible vectors, <strong>the</strong>n, given <strong>the</strong> value of <strong>the</strong> smallest<br />

component, <strong>the</strong> next smallest component is maximized, etc.<br />

In o<strong>the</strong>r words, we know that <strong>the</strong> total number of feasible vectors is <strong>in</strong> nite.<br />

For each allocation vector, <strong>the</strong> source that is gett<strong>in</strong>g <strong>the</strong> least allocation is <strong>in</strong> some<br />

sense, <strong>the</strong> \unhappiest source." Given <strong>the</strong> set of all feasible vectors, nd <strong>the</strong> vector<br />

that gives <strong>the</strong> maximum allocation to this unhappiest source. Actually, <strong>the</strong> number<br />

of such vectors is also <strong>in</strong> nite although we have narrowed down <strong>the</strong> search region<br />

considerably. Now we take this \unhappiest source" out and reduce <strong>the</strong> problem to<br />

that of rema<strong>in</strong><strong>in</strong>g n ; 1sourcesoperat<strong>in</strong>g on a network with reduced l<strong>in</strong>k capacities.<br />

Aga<strong>in</strong>, we nd <strong>the</strong> unhappiest source among <strong>the</strong>se n ; 1 sources, give that source<br />

<strong>the</strong> maximum allocation and reduce <strong>the</strong> problem by one source. We keep repeat<strong>in</strong>g<br />

this process until all sources have been given <strong>the</strong> maximum that <strong>the</strong>y could get. In<br />

summary,anetwork is considered to be <strong>in</strong> a state of max-m<strong>in</strong> fairness if it is impossible<br />

to <strong>in</strong>crease <strong>the</strong> rate of any session without decreas<strong>in</strong>g <strong>the</strong> rate of sessions whose rate<br />

is equal or smaller.<br />

Figure 3.3: Sample con guration <strong>for</strong> max-m<strong>in</strong> fairness<br />

43

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!