Re: [apops] using NTP/etc to measure n/w latency?
- To: apops at lists dot apnic dot net
- Subject: Re: [apops] using NTP/etc to measure n/w latency?
- From: Neale Banks <neale at lowendale dot com dot au>
- Date: Mon, 2 Oct 2000 19:54:44 +1100 (EST)
- Cc: "Henk Uijterwaal (RIPE-NCC)" <henk at ripe dot net>, Philip Smith <pfs at cisco dot com>
- In-reply-to: <Pine.BSI.4.05L.10010021011560.28322-100000 at x49 dot ripe dot net>
- Sender: email@example.com
Thanks for the suggestions so far. On Mon, 2 Oct 2000, Henk Uijterwaal (RIPE-NCC) wrote: > The (fundamental) problem with using NTP is that the accuracy of NTP > depends on the network latencies, which is also the quantity that you want > to measure. Yes, I still haven't wrapped my brain around how NTP gets around this apparent paradox. Reading over the RFCs-various hasn't helped me so far. Assuming that NTP does solve this problem, My logic is that if it reliably extracts the one-way latency (how else do you set your clock from the remote? ;-) then this latency is indeed the holy grail I seek. I'm fast concluding that ideally both ends would have some kind of external reference (e.g. each of "nearby" stratum-1 or, better still, their own non-network reference like GPS) for my purposes, but I would prefer to solve this with a minimum number of such references. Regards, Neale. * APOPS: Asia Pacific Operations Forum * * To unsubscribe: send "unsubscribe" to apops-request at apnic dot net *