Hi Guys,

I have a fair understanding of DSCP AFxx values.

Can someone please clarify the following for me:

1) For example, what's the difference between AF3x and CS3?

2) In what circumstances would we prefer to use CS3 instead of AF3x ?

Thanks
Aamir

The differences are mainly in how the application handles them. Because CS, IP precedence, AF values and DSCP all share the same bits of the TOS byte, it is up to the application to derive the significance.

CS3 shares many of the same bits as AF3x. You would choose one or the other based on that the application expects. If the application knows how to interpret AF31, then you would probably want to use that over CS3 because it is more granular in meaning.

But again, you choose your markings based on what the application can actually understand.

Hope that helps!

Thanks Blaag that was quite helpful :)

Blaag

Quote:

I disagree whit that because AF31 is allegible to drop if the queue is full, while CSXx is not. Cisco is going to change that in all the applications, the are using AF31 but are migrationg to CSXX.


Hope this helps.

hi Guys,
Both of these are diffrent way to classify streams for QOS purposes, CS is to provide backward compatibilty to the IP precedence based devices and works on the same pattern (the higher is better). The way it relates to IP prec., needs some explaining : (facts to rememeber CS is to provide backward compatibily to IP Prec)
IP precedence is 3 highest order bits of the TOS byte (bits 5-7) so all possible value can be 7 (bit one being 1, bit 2 being 2 and bit 3 being 4)

now when DSCP was introduced it took 6 higher order bits of the TOS byte and introduced four classes of services and 3 drop probabilities within each class (remember bigger is not always better)
and class selector are the same bit as IP pres.
for example AF31 will translate in to decimal value 26 ( the way to convert Af is 8*x+Y*2 (where X and Y are the first and second digits of the AF value)
so I was to put in to binary value mapping the DSCP :

32 16 8 4 2 1 (values from all 6 DSCP bits according to place value)
now, If I put 26 in to the DSCP value then:
0 1 1- 01 0 (i have put the hyphen to clear the boundary to indicate the IP prec. boundary)
so If a device only designed to read IP prec. value of qos streams will only read the first highest order bits and make the decision accroding to CS bits instead of DSCP.
another example:
voice is DSCP EF and value is 46 and dscp binary value would be
1 0 1 1 1 0
and hence recomended IP prec. value for voice is 5 if we only take first 3 highest order bits then
4 2 1 (place values of these bits)
then put the place value of
1 0 1 = 4+1 =5
thus providing compatibilty with IP prec.