MHonArc test archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Reply] Re: A question

Johns Daniel wrote:

> I would be interested in getting more information
> on how you measured CTD and CDV. I did this a
> while back using timestamps inserted by the app.
> into AAL5 packets. Unfortunately, I was not
> confident that my method was correct.

Well, I do it in a very simple way. To measure CTD I send cells from A
to  B and then again from B to A. A timestamps the output time and the
input time and dividing this by two you have an aproximation of the CTD.
For me this time is 232 usec. Due to the way timestamping is done:

	A				B

	           ------------------->	 read
                <-------------------     write	

this time doesn't mean anything. The time a socket read/write operation
takes (for 40 bytes, since the AAL5 layer inserts a 8 byte header, and I
just want to send 1 cell) is around 30 usec (sometimes it takes only 19
usec, but on average -- probably because other tasks are running at the
same time -- it takes 30 usec). Therefore the CTD discounting the
read/write overhead is 232 - 3 x 30 = 142 usec. From the application
point of view we have to add the time consumed by one read and one write
operation, hence 142 + 2 x 30 = 202 usec.

To compute the CDV I just send cells from A to B and timestamps the
elapsed time between the arrival of one cell and the arrival of the next
one. The problem with this method is that at high-speeds the read/write
operations become the bottleneck, and thus the times measured equal to
the time it takes to perform this operations (aprox. 32 usec).

A way to solve this problem is to reduce the speed, or maybe to use a
block of cells as a single cell for the simulations.

Hope it helped. If you need anything else (i.e. the code) don't hesitate
to contact me. I expect to produce more accurate results for next week.

Bye, Alfonso.

"Copy from one, it's plagiarism; Copy from two, it's reseach"

Alfonso Gonzalez 
Msc. High Speed Networks & Distributed Systems
Oxford Brookes University

Home | Main Index | Thread Index