Network cables - there are plenty of them in a data center. We don’t think about them too much. But do you remember that the type and length of your network cables impact your latency? We were recently reminded of this fact ourselves.
Minimizing latency wherever possible is our business, which is why we periodically verify the latency of our different offerings with physical hardware. When performing one such measurement, we noticed that the results were off by about 1.6 nanoseconds. We had to investigate further and figure out what was causing the issue.
We finally discovered the culprit: the server in question was connected to the switch with a copper cable that was 17 centimeters longer than we expected.
After some back and forth, we finally discovered the culprit: the server in question was connected to the switch with a copper cable that was 17 centimeters longer than we expected. This extra length added 1.6 nanoseconds to the round-trip latency, explaining the discrepancy in our measurement.
In case you need a reminder like us, 1 meter of copper cable adds about 4.60 ns, while for optical cable, it’s 4.96 ns per meter. Keep that in mind the next time you’re connecting your servers to the exchange.