Hello All,
I have noticed that the elapsed time is not equal to the sum of the user
and system times when using the tcp modules. For example, one might
record;
user time: 550s
sys time : 50s
real time: 750s
I first suspected that the missing time was the time in which the
communication actually happened; however, the opposite of that seems to be
happening.
In a job which transferes a total of:
4 gigabytes 1 gigabyte
user 580 650
sys 60 30
real 760 850
-----------------------------------
diff 120 170
The times are close, even though one program transfered one-fourth of the
data of the other. There may have been a difference in the size of the
packets, but they should have all been large.
Does anyone know a fairly good network profiler for 2.6.x kernels to look
into whats happening? Or does anyone know about this missing time right
off?
And what ever happened to M-VIA and VIA? They would help reduce processor
load some. Is it just that TCP can pump stuff out at near wire-speed as
it is, so there is no need for VIA? Has anyone ever thought of using
IPX/SPX networks? For clusters, they might be more efficient than TCP/IP.
------------------------------------------------------------
Anthony Ciani (aciani1_at_[hidden])
Computational Condensed Matter Physics
Department of Physics, University of Illinois, Chicago
http://ciani.phy.uic.edu/~tony
------------------------------------------------------------
|
