On 09/30/2010 08:48 AM, Kent A. Reed wrote:
>    On 9/29/2010 Jon Elson wrote:
>    
>> ontent-Type: text/plain; charset=ISO-8859-1; format=flowed
>>
>> Igor Chudov wrote:
>>      
>>>>   Things may have somewhat improved.
>>>>
>>>>   I checked on my unsophisticated home network. Ping time (roundtrip),
>>>>   involving three switches (one in my basement office, then the main
>>>>   switch at the main interconnect in the utility room, then the switch
>>>>   in the family room), and two linux boxes, is 0.21-0.34 milliseconds.
>>>>
>>>>          
>> My understanding of ping is that it does NOT report the total round trip
>> time through
>> all nodes and switches, just the last hop.  I think you need traceroute
>> to see the delay
>> at each hop.  Still, 300 uS is not such a great time if you need 3
>> messages to propagate
>> within one millisecond.
>>
>> Jon
>>
>>      
> Jon:
>
> Are you sure about that? I thought ping worked by sending an ICMP packet
> containing a time stamp and comparing that time stamp with its time of
> receipt of the returned packet. When I look at a test case I just did in
> my local network with ping and traceroute I see the same r/t times.
> Maybe my network just isn't complicated enough?
>
> In any case, ping and traceroute use ICMP packets and I don't believe
> they exercise the full protocol stack.
>
> Last year when I was so confidently saying I meant to try out RTnet with
> RTAI, I intended to be sure I tested the full time through the stacks,
> application to application. I started to sketch out minimal programs,
> one for the sender that, like ping, creates and sends a time-stamped
> packet and analyzes the packet returned, and a second for the client,
> that receives a packet and sends it back. A background program, like the
> EMC latency test, would track the reported times statistically.
> Unfortunately, I didn't have any boards with the requisite ethernet
> chips back then, and my attention wandered as my grandkids got older:-)
>
> Regards,
> Kent
>
> PS - thanks for setting me straight on the Windows/NT issue in the EMC
> project. I'd forgotten. I can salvage some of my dignity by noting the
> result adds to my argument that we have to demand functionality.
>    

Looking at the man page for ping, the "old" behavior of ping was to 
report "user-to-user" time.  The newer versions now report round-trip time.

Mark

------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing.
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to