Before you start reading, yes, I fully understand how silly this question is. But I need to give something to a customer who has the ability to run ping/traceroute but nothing else. (And they have an intermittent latency problem that we haven’t been able to isolate yet.)
Does anyone curate a list of “useful” ICMP responders that are at least kinda-sorta reliable/expected to continue responding? For example, all the major anycast DNS cloud providers respond to ICMP, but I don’t really want to tell my customer to ping an anycast IP address because the RTT results will be useless data (for comparative purposes).
I’m also not excited about providing random router IP address for what should be obvious reasons. There are some IPs that my routing paths that should be stable, but between routing changes and control-plane policing, those aren’t awesome. I’m looking for IPs I can suggest that are well outside my network.
Restatement: yes, there are much better ways to diagnose problems, but my customer can only run ping & traceroute (and pathping, I suppose) and is capable enough to run those tools and self-assess before calling me.
It sounds foolish to even ask, but maybe there’s a resource out there I don’t know about…
-Adam
Adam Thompson
Consultant, Infrastructure Services
![MERLIN]()
100 - 135 Innovation Drive
Winnipeg, MB, R3T 6A8
(204) 977-6824 or 1-800-430-6404 (MB only)
athompson@merlin.mb.ca
www.merlin.mb.ca
Before you start reading, yes, I fully understand how silly this question is. But I need to give something to a customer who has the ability to run ping/traceroute but nothing else. (And they have an intermittent latency problem that we haven’t been able to isolate yet.)
Over on this enormous thread, a gang of dedicated openwrt developers have been attempting to build an automated latency with load sensing tool leveraging sch_cake, currently targetted at LTE. They’ve released a few versions of the tool in lua and shell, so far.
Among other things, they’ve discovered that icmp type 13 (with a timestamp) actually works in many cases, and have a curated list mechanism, and
means to kick out and cope with various forms of responders.
Before you start reading, yes, I fully understand how silly this question is. But I need to give something to a customer who has the ability to run ping/traceroute but nothing else. (And they have an intermittent latency problem that we haven’t been able to isolate yet.)
Over on this enormous thread, a gang of dedicated openwrt developers have been attempting to build an automated latency with load sensing tool leveraging sch_cake, currently targetted at LTE. They’ve released a few versions of the tool in lua and shell, so far.
Among other things, they’ve discovered that icmp type 13 (with a timestamp) actually works in many cases, and have a curated list mechanism, and
means to kick out and cope with various forms of responders.
oops, the link:
https://forum.openwrt.org/t/cakes-autorate-ingress/108848/
Back in the old days, when there was competition between ISPs, the ISPs you paid money used to have curated speed test targets on their networks. Because you were paying them, some people wanted evidence they were getting what they were paying for, that was the only one under their control when you didn't get what you were paying.
"Testing" connectivity to random places on the internet, which depends on that wierd thing called peering, would give you random results which no one would do anything about.
Step 1: Ask the ISP, which you are paying, for appropriate targets.
Does anyone curate a list of “useful” ICMP responders that are at least kinda-sorta reliable/expected to continue responding?
RIPE Atlas anchors are useful for this kind of thing
https://atlas.ripe.net/anchors/list/
We ran speedtest.net servers for about 4 years, and then retired them.
They outlived their usefuless as the minimum bandwidth made available to customers started to test the limits both of the servers and the customer's own gear.
It took another 3 years for both customers and internal teams to get onboard with "no more speedtest.net here", but finally, seems the ship has turned.
Mark.
We do not recommend or advise customers to ping random online resources, however famous they may be. We don’t want to be part of growing that problem for other networks. We do operate a number of FreeBSD servers to capture and share public telemetry, and do offer customers the option to ping those, in various cities, if they are really keen to. Sometimes they do, other times they prefer to ping the famous online addresses. We only put our name behind the servers we operate. We do not stop them from pinging off-net resources, but we make no gurantees as to their experience doing that. Mark.
We do not recommend or advise customers to ping random online
resources, however famous they may be. We don't want to be part of
growing that problem for other networks.
i helped the shy admin of a well known site talk to an entire
politically sensitive country which had decided to ping that site to
test liveness/access
but, heck, we don't publish a list of pingables. so we're gonna get
random behavior. and some days all the air molecules go to one corner
of the room.
randy
Fair point.
Most of our customers are not interested in our on-net servers that they can ping. They prefer to ping the usual suspects, because, well, that is "the Internet", to them.
Mark.