- Newest
- Most votes
- Most comments
What does "high" mean? How is "high" different from what is expected? There are no details in the question about what services are being used; what the latency is; when the impact is happening; what the application is doing; or what you expect the result to be.
If this is a production issue, please raise a support case. Otherwise, please put some additional details here which makes it easier to provide additional advice.
Edit (adding here because it's easier to read): TL;DR: The speed of light is finite and increases in latency over large distances are to be expected.
If we assume the ideal case: Given that Cape Town is approximately 9,300 kms from Paris then light in a vacuum will take about 31 milliseconds to travel that distance, giving us a round-trip time of 62 milliseconds. But the speed of light in fibre optic is about 30% less; and there are probably no fibre optic cables running in a straight line from Paris to Cape Town so even if we assume that the round-trip time is about 88 ms it's probably closer to 100 ms. But let's use 88 ms as a round-trip latency and 44 ms as the one-way anyway just to see what happens.
What we don't know here is how many packets are being exchanged to do the API call Let's take a guess and say it is 30 and let's ignore any latency introduced by the service itself although that's going to be a factor - but it is a constant factor.
So for a 115 ms API call in Paris with 30 packets the latency is going to be just under 4 ms which seems reasonable so 30 is a good number to pick.
At 44 ms, 30 packets take 1320 ms which is pretty much spot on with what you're seeing. It should be higher but I'm ignoring a bunch of other factors here (it could be less packets; there are definitely optimisations that happen at the TCP layer; and so on). But what you're seeing is (in my mind) about right.
Hi,
I'd suggest to implement first AWS Network Monitor to be able to come with precise numbers in order to solid assistance from re:Post or AWS Support
See https://aws.amazon.com/cloudwatch/features/network-monitoring/
The right way would be to provide comparative numbers between different pairs of regions showing the difference South Africa <> Paris vs other pairs.
Best,
Didier
Relevant content
- asked 2 years ago
- asked a year ago
- AWS OFFICIALUpdated 3 years ago
I hope this message finds you well.
We have been monitoring the performance of the GetLanguage API hosted in the Paris region (API Gateway). During a recent review, we observed a significant discrepancy in the response times when the API is called from different regions.
Specifically, when the GetLanguage API is called from Cape Town to Paris, the response time was recorded at 1.391 seconds. However, when the API is called within the same region (Paris to Paris), the response time dropped significantly to just 115 milliseconds.
At this point, we have not been able to pinpoint the exact cause of this performance drop or whether this is due to a network issue or another underlying factor. We will continue investigating further, but your insights or assistance in identifying the root cause would be highly appreciated.
Please let us know if there are additional diagnostics we should consider or any other information you might need from our end.
Thank you for your attention to this matter.