Need host with very large ping times

I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter). Amazingly, I could not easily find one. I even tried to ping-scan IP ranges of Antarctica, but those hosts that responded did so in fraction of a second. Not much luck with North Korea (mostly down), Taiwan (internet is much better than it used to be) and several other suspect countries. (Though I'm not sure my tests are correct, since firewall processing of such long response times is one of the things I'm trying to find out.) Is it possible to somehow find one using RIPE ATLAS, preferably without ping-scanning whole internet? Would be especially great if it responded to other types of ping beside ICMP. -- With Best Regards, Marat Khalili

On 26 May 2017, at 13:51, Marat Khalili <mkh@rqc.ru> wrote:
I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter). Amazingly, I could not easily find one. I even tried to ping-scan IP ranges of Antarctica, but those hosts that responded did so in fraction of a second. Not much luck with North Korea (mostly down), Taiwan (internet is much better than it used to be) and several other suspect countries. (Though I'm not sure my tests are correct, since firewall processing of such long response times is one of the things I'm trying to find out.)
Is it possible to somehow find one using RIPE ATLAS, preferably without ping-scanning whole internet? Would be especially great if it responded to other types of ping beside ICMP.
I haven’t tried such long delays, but FreeBSD’s dummynet allows you to create pipes with arbitrarily long delays, percentage of packet loss, etc Best regards, Borja.

You could try to test with this tool: https://github.com/tylertreat/comcast On 26.05.17 13:51, Marat Khalili wrote:
I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter). Amazingly, I could not easily find one. I even tried to ping-scan IP ranges of Antarctica, but those hosts that responded did so in fraction of a second. Not much luck with North Korea (mostly down), Taiwan (internet is much better than it used to be) and several other suspect countries. (Though I'm not sure my tests are correct, since firewall processing of such long response times is one of the things I'm trying to find out.)
Is it possible to somehow find one using RIPE ATLAS, preferably without ping-scanning whole internet? Would be especially great if it responded to other types of ping beside ICMP.
--
With Best Regards, Marat Khalili

Annika, Borja, thank you for the suggestions. I'll consider purpose-building such host, but still hope to stumble upon existing one (some remote observatory on a satellite link probably...) -- With Best Regards, Marat Khalili On 26/05/17 14:57, Annika Wickert wrote:
You could try to test with this tool: https://github.com/tylertreat/comcast
On 26.05.17 13:51, Marat Khalili wrote:
I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter). Amazingly, I could not easily find one. I even tried to ping-scan IP ranges of Antarctica, but those hosts that responded did so in fraction of a second. Not much luck with North Korea (mostly down), Taiwan (internet is much better than it used to be) and several other suspect countries. (Though I'm not sure my tests are correct, since firewall processing of such long response times is one of the things I'm trying to find out.)
Is it possible to somehow find one using RIPE ATLAS, preferably without ping-scanning whole internet? Would be especially great if it responded to other types of ping beside ICMP.
--
With Best Regards, Marat Khalili

Hi Marat, there are big datasets of results of people that did already ICMP scans of the Internet, that contain timestamps, from which you could extract hosts with large RTTs. This is a scan from Univ. Michigan (as i understand it): https://censys.io/data/0-icmp-echo_request-full_ipv4 This research paper: http://conferences2.sigcomm.org/imc/2015/papers/p303.pdf describes how they found IP addresses with high RTTs (5% of IP address space that was probed had >5s RTT for 5% of the time). Maybe you can contact the authors if you are interested in this data? hope this helps, Emile Aben RIPE NCC On 26/05/17 14:17, Marat Khalili wrote:
Annika, Borja, thank you for the suggestions. I'll consider purpose-building such host, but still hope to stumble upon existing one (some remote observatory on a satellite link probably...)
--
With Best Regards, Marat Khalili
On 26/05/17 14:57, Annika Wickert wrote:
You could try to test with this tool: https://github.com/tylertreat/comcast
On 26.05.17 13:51, Marat Khalili wrote:
I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter). Amazingly, I could not easily find one. I even tried to ping-scan IP ranges of Antarctica, but those hosts that responded did so in fraction of a second. Not much luck with North Korea (mostly down), Taiwan (internet is much better than it used to be) and several other suspect countries. (Though I'm not sure my tests are correct, since firewall processing of such long response times is one of the things I'm trying to find out.)
Is it possible to somehow find one using RIPE ATLAS, preferably without ping-scanning whole internet? Would be especially great if it responded to other types of ping beside ICMP.
--
With Best Regards, Marat Khalili

Very interesting research indeed! Section "ASes most prone to RTTs greater than 1 second" contained enough information for me to find these hosts, no need to contact authors. Quest complete, thank you! -- With Best Regards, Marat Khalili On 26/05/17 16:32, Emile Aben wrote:
Hi Marat,
there are big datasets of results of people that did already ICMP scans of the Internet, that contain timestamps, from which you could extract hosts with large RTTs. This is a scan from Univ. Michigan (as i understand it): https://censys.io/data/0-icmp-echo_request-full_ipv4
This research paper: http://conferences2.sigcomm.org/imc/2015/papers/p303.pdf describes how they found IP addresses with high RTTs (5% of IP address space that was probed had >5s RTT for 5% of the time). Maybe you can contact the authors if you are interested in this data?
hope this helps, Emile Aben RIPE NCC
On 26/05/17 14:17, Marat Khalili wrote:
Annika, Borja, thank you for the suggestions. I'll consider purpose-building such host, but still hope to stumble upon existing one (some remote observatory on a satellite link probably...)
--
With Best Regards, Marat Khalili
On 26/05/17 14:57, Annika Wickert wrote:
You could try to test with this tool: https://github.com/tylertreat/comcast
On 26.05.17 13:51, Marat Khalili wrote:
I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter). Amazingly, I could not easily find one. I even tried to ping-scan IP ranges of Antarctica, but those hosts that responded did so in fraction of a second. Not much luck with North Korea (mostly down), Taiwan (internet is much better than it used to be) and several other suspect countries. (Though I'm not sure my tests are correct, since firewall processing of such long response times is one of the things I'm trying to find out.)
Is it possible to somehow find one using RIPE ATLAS, preferably without ping-scanning whole internet? Would be especially great if it responded to other types of ping beside ICMP.
--
With Best Regards, Marat Khalili

On 26 May 2017, at 14:17, Marat Khalili <mkh@rqc.ru> wrote:
Annika, Borja, thank you for the suggestions. I'll consider purpose-building such host, but still hope to stumble upon existing one (some remote observatory on a satellite link probably…)
An example creating a 20 second delay. # ipfw pipe 10 config delay 9999 Now we create a couple of firewall rules. I will give this “royal treatment” to betweeen “pinger” and “pingee” # ipfw add 121 pipe 10 icmp from x.y.z.t to me # ipfw add 122 pipe 10 icmp from me to x.y.z.t (Adding delay to both the ping and the ping ack so that the maximum delay I can get is 20000 ms) And testing: 64 bytes from a.b.c.d: icmp_seq=259 ttl=60 time=20000.069 ms If you need it, give me the IP addresses from which you intend to do the tests and the delay you need and I can create a ping target on my test server in Amazon’s EC2. Cheers, Borja.

Dear Borja, Thank you for your offer, but I already found out what I wanted for fixed source IP address and several seconds delays (that firewalls are ok with it). For "real" research much more would need to be done, but I don't have enough time and resources for it now. Just in case someone is interested in conducting this kind of research, IMO it would be interesting to open delay host to all internet and point ATLAS probes to it, then vary delay time and try to find out delays at which firewalls start to close connections prematurely. Then repeat it for TCP and (especially interesting) UDP, but AFAIU current probes can only send TCP/UDP test packets to anchors, therefore it would require coordination with ATLAS people. (I did not study literature, so maybe it has already been done recently.) -- With Best Regards, Marat Khalili On 29/05/17 09:47, Borja Marcos wrote:
On 26 May 2017, at 14:17, Marat Khalili <mkh@rqc.ru> wrote:
Annika, Borja, thank you for the suggestions. I'll consider purpose-building such host, but still hope to stumble upon existing one (some remote observatory on a satellite link probably…) An example creating a 20 second delay.
# ipfw pipe 10 config delay 9999
Now we create a couple of firewall rules. I will give this “royal treatment” to betweeen “pinger” and “pingee”
# ipfw add 121 pipe 10 icmp from x.y.z.t to me # ipfw add 122 pipe 10 icmp from me to x.y.z.t
(Adding delay to both the ping and the ping ack so that the maximum delay I can get is 20000 ms)
And testing:
64 bytes from a.b.c.d: icmp_seq=259 ttl=60 time=20000.069 ms
If you need it, give me the IP addresses from which you intend to do the tests and the delay you need and I can create a ping target on my test server in Amazon’s EC2.
Cheers,
Borja.

On Fri, 26 May 2017 14:51:49 +0300 Marat Khalili <mkh@rqc.ru> wrote:
I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter).
Just something to keep in mind, if they have such a terrible network connection, they probably would not appreciate being pinged (if you mean continuously or at any degree of regularity), chipping away at their miniscule network bandwidth in favor of some random person's unrelated research... -- With respect, Roman

On May 26, 2017, at 12:42 PM, Roman Mamedov <rm@romanrm.net> wrote:
On Fri, 26 May 2017 14:51:49 +0300 Marat Khalili <mkh@rqc.ru> wrote:
I have an interesting problem: I need some host that doesn't mind being pinged but with several seconds of RTT (that's thousands milliseconds; from Moscow, but it probably doesn't matter).
Just something to keep in mind, if they have such a terrible network connection, they probably would not appreciate being pinged (if you mean continuously or at any degree of regularity), chipping away at their miniscule network bandwidth in favor of some random person's unrelated research...
My experience is it’s not always about the lack of connectivity, it’s about odd buffering behaviors of devices. The Juniper M40 would buffer an incredible amount of data on the FPC for a T1 speed link whereas a OC48 FPC had the same buffer sizes for much higher rates. Of course, this all assumes you know what you’re measuring, which is a critical part of it. - Jared
participants (6)
-
Annika Wickert
-
Borja Marcos
-
Emile Aben
-
Jared Mauch
-
Marat Khalili
-
Roman Mamedov