How does Skew impact latency measurement?


Now more than ever latency is a key network quality metric, with financial institutions and other Enterprise customers demanding absolutely minimal latency from their service provider.

With Skew being an inherent part of a 100G Ethernet link, field engineers starting to roll out these services commonly ask how will Skew levels affect the latency measurements they make when perform service turn up testing.

Per the IEEE 802.3ba document the maximum amount of Skew within a received PCS lane shall be 928 bits. It will be present within 19 of the 20 PCS lanes (with the remaining lane being the first lane received at the PCS). Therefore when we calculate the effect of Skew on end to end latency we must take into account that we must calculate this based on the data rate of a PCS lane (5.1625 Gbaud) and not the total interface rate.

At 5.1625Gbaud the effect of a 928 bit delay represents a maximum additional latency of only 1.8*10-7s (0.18µs). Over a complete circuit this additional delay may be found at any point where there is a 100G Base-R link.

Overall the effect of Skew on the latency can be considered negligible and unlikely to have any measurable impact on the end to end service testing. It would only be necessary to measure the Skew in order to verify that it is not at such a level where it would affect the Ethernet link.

Leave a Reply