& cplSiteName &
Comments
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
Sr.Embed22197
50%
50%
Sr.Embed22197,
User Rank: Light Beer
5/15/2017 | 2:18:37 PM
Re: Ummm
It's actually a bit of both, yes.

I guess my original point is that One Way measurments have some clock sync requirements that are far more strict than Two Way.  Y.1731 DMM, TWAMP, the test can compensate to a certain degree because you can factor out the time differences at each end.  One Way testing, the clocks have to be sync'd a lot tighter.

It's a price upgrade for hardware and infrastructure to do that and customers hate spending the extra money.  Most switches currently installed that I have to support, still rely on NTP.

The point of the article is that One Way testing is needed, and I'm seeing requests to add it, as well as defects reports about it's accuracy on older devices.  I agree that it's useful, but making it report valid numbers isn't just a slight tweak of the software. :)
brooks7
50%
50%
brooks7,
User Rank: Light Sabre
5/15/2017 | 1:33:46 PM
Ummm
I think about 1/2 your commentary is actually about Jitter and not latency.  Absolute Latency is important at the start of a stream (which includes channel switching).  But once a stream has started it is packet jitter that is the problem (i.e. the variance in delay between receipt of packets).

seven

 
Sr.Embed22197
50%
50%
Sr.Embed22197,
User Rank: Light Beer
5/15/2017 | 12:50:48 PM
Re: When the devices clocks are accurate enough ...
I'm working with 100G Ethernet. :)  PtP is the protocol they should be using, to have a chance of being accurate, so I agree with you there.  It's not available on low end switches though and customers want cheap switches.

Over multiple hops, from the customers testing center to the NTE, the end to end clocks have be be in sync if you're going to use something like Y.1731 1DM.  Sure, there are games you can play, run 100 frame tests, throw out the outliers, take the average.  You're still going to get odd results at times because someplace in the middle of the test, NTP updated the clock and slewed it someplace.

Customers hate running repeating tests and noticing occasional results that wandered out into the weeds because of a clock update.  Everytime a leap second hits, we get a couple of calls and we explain it all over again.
Duh!
50%
50%
Duh!,
User Rank: Blogger
5/15/2017 | 12:29:14 PM
Re: When the devices clocks are accurate enough ...
Off by a couple of orders of magnitude. Prop delay through fiber is 4.9 microseconds (μs) per kilometer. Serialization delay at 1 Gbps for an MTU-sized packet is 12.3 μs.

IEEE 1588 - Precision Time Protocol can provide commensurate accuracy.
Sr.Embed22197
50%
50%
Sr.Embed22197,
User Rank: Light Beer
5/15/2017 | 12:07:11 PM
When the devices clocks are accurate enough ...
When most vendors/customers use NTP to keep their clock accurate, I'd love to hear somebody explain how a clock that is +/- 128ms accuracy is going to get used to time the one way flight of a packet that takes nanoseconds to travel.

Yeah, there are better time protocols and sync Ethernet, but nobody wants to pay for it.

 


Featured Video
Upcoming Live Events
September 17-19, 2019, Dallas, Texas
October 1-2, 2019, New Orleans, Louisiana
October 10, 2019, New York, New York
October 22, 2019, Los Angeles, CA
November 5, 2019, London, England
November 7, 2019, London, UK
November 14, 2019, Maritim Hotel, Berlin
December 3-5, 2019, Vienna, Austria
December 3, 2019, New York, New York
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events