Rate Measurement!! Contd...

[Eddie Kohler]

> Not in the cases that I had shown earlier, but in my
> implementation, I am using PollDevice(eth0) to receive packets ->
> ->  Shape at a rate 'R' ->  ToDevice(eth1).
>                 Do you think that in my implementation I can be assured that
> the
> rate at which packet is sent out is 'R' bytes/second. What are the ways in
> which PollDevice
> could be a contributing factor in limiting 'R'?

It depends on R. If you set R = 1000000000000000000000000000000000000
packets/sec, then obviously you are not assured that rate :)

Moving packets through a PC stresses many parts of that machine. Device
hardware, device drivers, memory subsystem, PCI bus arbitration and
bandwidth, CPU, .... Any of these can 'max out', limiting your rate; from a
Click viewpoint, they can all look like PollDevice is limiting the rate. My
thesis, available off the Click homepage, has a chapter detailing many of
these issues that you might want to look at.

Eddie