&y0 <> wrote on Friday, January 28, 2005 12:52 PM:
> Hi everybody. I have problem while customers places calls to 5350 NAS
> via ISDN using ISDN-TA and PPP-multilink. During the call 2
> B-channels are bound to Virtual-access interface which parameters
> then looks like this:
>[...]
> Accordingly customer can obtain the data transfer speed no more ~88K.
> It looks like 16K is spent for unknown reason per each B-channel as
> arithmetic of 2*(64K-16K)=96K shows I have ever trusted that
> signalling bandwidth should be taken from D-channel bandwidth (in my
> case Serial3/0:15).
Can you check if you can saturate the link (i.e. full 128k) using
multiple TCP sessions? TCP throughput is goverend by many parameters,
link delay being an important one. Or try using a blast ping or
something which just fills the link not worrying about any
acknowledgements..
oli
> Hi everybody. I have problem while customers places calls to 5350 NAS
> via ISDN using ISDN-TA and PPP-multilink. During the call 2
> B-channels are bound to Virtual-access interface which parameters
> then looks like this:
>[...]
> Accordingly customer can obtain the data transfer speed no more ~88K.
> It looks like 16K is spent for unknown reason per each B-channel as
> arithmetic of 2*(64K-16K)=96K shows I have ever trusted that
> signalling bandwidth should be taken from D-channel bandwidth (in my
> case Serial3/0:15).
Can you check if you can saturate the link (i.e. full 128k) using
multiple TCP sessions? TCP throughput is goverend by many parameters,
link delay being an important one. Or try using a blast ping or
something which just fills the link not worrying about any
acknowledgements..
oli