Siobhan P. Lynch wrote: > What I'm trying to do is get timestamps on echo packets going out and > the timestamp on a reply coming in, very specifically, I know what the > client thinks he's looking at, what I'm trying to prove is the issue is > that he's putting 30 lbs of garbage in a 20 lb. bag (squeezing more data > down a 144k IDSL line than they should be)
hmm ... that sounds a lot more specific than "I can't see what he's talking about". I'm optimistic that with the detail you gave, more help will be forthcoming. Please tell us what build of opensolaris you're using. Have you tried snoop? It's not DTrace, but it's very good for observing network traffic, and *does* include timestamps (or so I recall). maybe [EMAIL PROTECTED] is also a good place to ask for suggestions about analysing this. HTH Michael > The issue, without getting too much into it is that they are seeing > extremely high icmp echo reply times through the box, they are blaiming > it on the box itself, so what I'm trying to do is establish a baseline. > > I'm not looking for "the answer" from you, I'm looking for examples > using DTrace on how to measure the exact times between two packets (one > request, one reply, icmp or tcp ping) essentially RTT times. > > It seems that anyone I ask doesn't know how to use the network provider > to do something like this, which is why I ask, I really have no examples > to go in in this particular situation, so I can't even "teach myself" > like I usually do. > > -Trish > >> -----Original Message----- >> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] >> Sent: Tuesday, March 04, 2008 10:22 AM >> To: Siobhan P. Lynch >> Cc: [email protected] >> Subject: Re: [dtrace-discuss] Network Latency >> >> Siobhan P. Lynch wrote: >>> Hiya, >>> >>> >>> >>> I'm trying to track down some throughput latency that our customer > seems >>> to be attributing to our product, I can't see what he's talking > about, >> [...] >> >> Please excuse if this is not what you want to hear: >> >> The *first* thing you have to do to solve a problem is to understand > it. >> You might be "getting some deeper granularity" only to find out after >> having spent some time on that that your customer *actually* means >> something slightly different than you do when you speak of "throughput >> latency". >> >> BTW: What is "throughput latency" supposed to mean? in my experience, >> throuput and latency are orthogonal concepts: throughput being the > average >> amount of data you get over a connection over time, and latency being > the >> response time to a given request. >> Have your customer clarify that first. >> >> HTH >> Michael >> -- >> Michael Schuster http://blogs.sun.com/recursion >> Recursion, n.: see 'Recursion' > -- Michael Schuster Sun Microsystems, Inc. recursion, n: see 'recursion' _______________________________________________ dtrace-discuss mailing list [email protected]
