Unformatted text preview: fore the updating. We use
b
an analogous formula to update F .
The updating rule for b depends on whether the link is
congested or not. To lter out the estimation inaccuracies
due to exponential smoothing we use a window of size Kc .
b
A link is assumed to be congested, if A C at all times during an interval of length Kc. Conversely, a link is assumed
b
to be uncongested, if A C at all times during an interval
of length Kc . The value b is updated only at the end of an
interval in which the link is either congested or uncongested
according to these de nitions. If the link is congested then
b is updated based on the equation F b = C . We approximate F by a linear function that intersects the origin and
b
has slope F= bold . This yields
C
bnew = bold
6
b F If the link is not congested, bnew is set to the largest rate
of any active ow i.e., the largest label seen during the
last Kc time units. The value of bnew is then used to compute dropping probabilities, according to Eq. 2. For completeness, we give the pseudocode of the CSFQ algorithm in
Figure 2.
We now describe two minor amendments to this algorithm related to how the bu ers are managed. The goal of
estimating the fair share b is to match the accepted rate to
the link bandwidth. Due to estimation inaccuracies, load
uctuations between b's updates, and the probabilistic nature of our algorithm, the accepted rate may occasionally
exceed the link capacity. While ideally the router's bu ers
can accommodate the extra packets, occasionally the router
may be forced to drop the incoming packet due to lack of
bu er space. Since droptail behavior will defeat the purpose
of our algorithm, and may exhibit undesirable properties in
the case of adaptive ows such as TCP 9 , it is important
to limit its e ect. To do so, we use a simple heuristic: every on receiving packet p
if edge router
i =classifyp;
p:label = estimate rateri ; p; = use Eq. 3 =
prob =max0; 1 , =p:label;
if prob unif rand0, 1
=estimate p; 1;
dropp;
else
=estimate p; 0;
en...
View
Full Document
 Fall '10
 Eugene Ng
 Scheduling algorithm, Roundrobin scheduling, Scheduling algorithms, Ow, Fair queuing, ows

Click to edit the document details