Is a difference of 0.6ms in ping times going to have a measurable impact to data transfer?

-1

I'm benchmarking our app across the various desktop pc's we have in our organisation. Our application is network-heavy (all processing is done server-side). I'm trying to interpret how much of the application's perceived performance is due to cpu vs network usage.

From our dataserver to 2 desktop machines I'm seeing quite different ping times and traceroute shows dropped packets. Are the differences in ping respons something that I need to be concerned about or should I ignore them and focus on the other hardware differences between my test machines?

Output (anonymized, but the route is identical to both target machines)

traceroute to testmachine1 (10.0.0.1), 30 hops max, 60 byte packets
 1  somehost1 (10.1.1.1)  1.315 ms  1.328 ms  1.419 ms
 2  somehost2 (10.1.1.2)  0.506 ms  0.556 ms  0.677 ms
 3  somehost3 (10.1.1.3)  0.494 ms  0.492 ms  0.489 ms
 4  testmachine1 (10.0.0.1)  0.658 ms  0.771 ms  0.770 ms

traceroute to testmachine2 (10.0.0.2), 30 hops max, 60 byte packets
 1  somehost1 (10.1.1.1)  0.805 ms  0.804 ms  0.875 ms
 2  somehost2 (10.1.1.2)  0.782 ms  0.867 ms  0.910 ms
 3  somehost3 (10.1.1.3)  0.569 ms  0.570 ms  0.571 ms
 4  testmachine2 (10.0.0.2)  1.234 ms * *

Test machine Specs

Drivers are up to date on both machines
Both running Windows 7 enterprise
Physical connections go identically through Avaya IP Phones
testmachine1 - HP Z620 - 2 x Xeon E5-2620, 12GB, 1GB onboard Intel NIC
testmachine2 - HP Z600 - 2 x Xeon E5506, 24GB, 1GB onboard Broadcom NIC

beirtipol

Posted 2015-12-07T14:49:59.280

Reputation: 255

I seem to have gotten 2 downvotes without any explanation?! – beirtipol – 2015-12-07T15:18:39.227

I didn't downvote, but I'm guessing that it may be because your question is very broad in general, and (especially 3rd bullet, possibly 2nd), is likely to attract answers based on opinion rather than facts. The site is designed to build a knowledgebase, so ideal questions are those that have a fact-based answer (i.e., something someone could look up in a reference source). Troubleshooting questions are hard to answer in this format, and often have little value to people other than the OP because they are so specific to conditions. – fixer1234 – 2015-12-07T16:30:54.260

Gotcha - thanks. I'm stuck trying to diagnose a problem with too many variables and hoping that I can knock some off by someone saying "as long as the ping time is less than 5 millis, you need not worry". I'll try and reword – beirtipol – 2015-12-07T17:39:10.723

Are you saying that even though the same route is used from the server to the 2 test machines, that the difference could be down to the router? Or that the routers could provide further information on what my test machines are doing? – beirtipol – 2015-12-07T17:43:40.353

Both actually. For example: the test-machine with less performance could be using a damaged port, or the cable used could be damaged in a way that it is still working but not perfectly. The router certainly does have more information specially if it is pro-grade or has custom firmware installed. Better even it has an SNMP server so you can use a client to check its status on-the-fly. – FernanDK – 2015-12-07T18:20:11.093

One more thing. You said that the application is network-heavy right.. If I had to take a blind shot I would say the difference you are seeing in ping are due to the hardware difference between the two machines.

testmachine1 has a way better CPU than testmachine2, more cache, more speed, more protocols, less heat, etc. It is also deducible that testmachine1 packs a more robust NIC since it is Intel. Note that each of this hardware differences are sufficient enough to cause those differences shown in traceroute, never mind together. – FernanDK – 2015-12-07T18:33:49.633

Yeah, far too many variables, but thank you for the points, much appreciated. – beirtipol – 2015-12-08T08:19:49.550

Is a difference of 0.6ms in ping times going to have a measurable impact to data transfer?

Answers