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.

m<strong>in</strong>imum fairshare is low. This may result <strong>in</strong> slower transient response,<br />

but does not result <strong>in</strong> overallocation.<br />

The number of active sources can uctuate if some sources are not seen <strong>in</strong><br />

an <strong>in</strong>terval. Fur<strong>the</strong>r, due to <strong>the</strong> cluster<strong>in</strong>g e ect of TCP, cells from just a<br />

few VCs may be seen <strong>in</strong> an <strong>in</strong>terval lead<strong>in</strong>g to an underestimate of Na.<br />

In averag<strong>in</strong>g Na, <strong>the</strong> scheme ma<strong>in</strong>ta<strong>in</strong>s an activity level <strong>for</strong> each source.<br />

The activity level of <strong>the</strong> source is set to one when any cell of <strong>the</strong> source<br />

is seen <strong>in</strong> <strong>the</strong> <strong>in</strong>terval. However, when no cell from a source is seen <strong>in</strong> an<br />

<strong>in</strong>terval, <strong>the</strong> scheme \decays" <strong>the</strong> activity level of <strong>the</strong> source by a factor,<br />

\ n"(also called DecayF actor). Hence, <strong>the</strong> source becomes <strong>in</strong>active only<br />

after many <strong>in</strong>tervals. A recommended value of n is 0.9. Roughly, <strong>the</strong> Na<br />

measured with this value of n is approximately equal to <strong>the</strong> Na measured<br />

without averag<strong>in</strong>g over an averag<strong>in</strong>g <strong>in</strong>terval 8 or 9 times larger than <strong>the</strong><br />

current averag<strong>in</strong>g <strong>in</strong>terval.<br />

3. Third, we modify <strong>the</strong> response to boundary conditions of <strong>the</strong> scheme. This<br />

allows <strong>the</strong> scheme to handle <strong>the</strong> boundary conditions gracefully. Speci cally,<br />

<strong>the</strong> number of active sources is set to one if it is measured to be below one.<br />

The second method of overload factor averag<strong>in</strong>g does not allow <strong>the</strong> overload<br />

factor be zero or <strong>in</strong> nity. However, outlier measurements are not ignored <strong>in</strong> <strong>the</strong><br />

averag<strong>in</strong>g method.<br />

The ERICA+ scheme with <strong>the</strong>se modi cations controls <strong>the</strong> <strong>ABR</strong> queues without<br />

overly compromis<strong>in</strong>g on TCP throughput. Table 8.9 shows <strong>the</strong> results of representa-<br />

tive experiments us<strong>in</strong>g <strong>the</strong>se features.<br />

310

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

Saved successfully!

Ooh no, something went wrong!