RIPE Atlas for Saudi Arabia - how come so many abandoned?
Hi all, I have a basic question. If I look at the latest stats for Saudi Arabia and I see: https://atlas.ripe.net/results/maps/network-coverage/?filter=Saudi+Arabia+(s...)
Connected: 8 Disconnected: 4 Abandoned: 43
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online? I could easily pick another country; however, I waned to do a measurement in KSA today, hence this is on my mind. Yours, Martin Martin J. Levy <martin@cloudflare.com> Distinguished Engineer @ Cloudflare, Inc. +1 408 499 3801 (mobile) +44 7852 156845 (when in UK) https://cloudflare.com/ (web)
On Apr 29, 2019, at 16:39, Martin Levy <martin@cloudflare.com> wrote:
What part can an ambassador play in order to get abandoned probes back online?
I thought I was good an encouraging participation ... until I tried the Atlas Ambassador program. It is “pushing rope.” What I find is that the people who use Atlas, program it into their tools, then have a motivation to keep their probes functional (they need the credits).
Hi Martin,
On 30 Apr 2019, at 00:39, Martin Levy <martin@cloudflare.com> wrote:
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online?
This is what happens when probes are distributed in the wild. In Africa, we have the same problem, now our strategy is to give it to interested collaborators only. It might be useful to understand the measurement needs of the host, some are interested in bandwidth, others in latency, DNS, etc. So providing them a probe + some pre-cooked visuals of the state of their network would be a good recipe imo. — Amreesh
Dear people from RIPE who are involved in Atlas project, do you have any past experience to get people (who have abandoned probes) notified over email with nice message asking to connect probes back (or return to ambassador for re-distribution)? -- Evgeniy On Tue, Apr 30, 2019 at 2:10 AM Amreesh Phokeer <amreesh@afrinic.net> wrote:
Hi Martin,
On 30 Apr 2019, at 00:39, Martin Levy <martin@cloudflare.com> wrote:
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online?
This is what happens when probes are distributed in the wild. In Africa, we have the same problem, now our strategy is to give it to interested collaborators only.
It might be useful to understand the measurement needs of the host, some are interested in bandwidth, others in latency, DNS, etc. So providing them a probe + some pre-cooked visuals of the state of their network would be a good recipe imo.
— Amreesh _______________________________________________ RIPE-Atlas-Ambassadors mailing list RIPE-Atlas-Ambassadors@ripe.net https://lists.ripe.net/mailman/listinfo/ripe-atlas-ambassadors
-- -- With regards, Eugene Sudyr
Good Morning Evgeniy, Yes, we do indeed have these challenges in Saudi and other some countries for different reasons. I contacted many of the concerned operators and ISPs informing them that they/we can’t do the measurements if the probes are not connected. I mentioned that our resources are limited, so we truly hope we can get the probes up to do full diagnose and measurements, or else we will have to release the resources that were assigned to them to other projects. Best Regards, Chafic
On 30 Apr 2019, at 09:05, Evgeniy S. <eject.in.ua@gmail.com> wrote:
Dear people from RIPE who are involved in Atlas project,
do you have any past experience to get people (who have abandoned probes) notified over email with nice message asking to connect probes back (or return to ambassador for re-distribution)?
-- Evgeniy
On Tue, Apr 30, 2019 at 2:10 AM Amreesh Phokeer <amreesh@afrinic.net <mailto:amreesh@afrinic.net>> wrote: Hi Martin,
On 30 Apr 2019, at 00:39, Martin Levy <martin@cloudflare.com <mailto:martin@cloudflare.com>> wrote:
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online?
This is what happens when probes are distributed in the wild. In Africa, we have the same problem, now our strategy is to give it to interested collaborators only.
It might be useful to understand the measurement needs of the host, some are interested in bandwidth, others in latency, DNS, etc. So providing them a probe + some pre-cooked visuals of the state of their network would be a good recipe imo.
— Amreesh _______________________________________________ RIPE-Atlas-Ambassadors mailing list RIPE-Atlas-Ambassadors@ripe.net <mailto:RIPE-Atlas-Ambassadors@ripe.net> https://lists.ripe.net/mailman/listinfo/ripe-atlas-ambassadors <https://lists.ripe.net/mailman/listinfo/ripe-atlas-ambassadors>
-- -- With regards, Eugene Sudyr _______________________________________________ RIPE-Atlas-Ambassadors mailing list RIPE-Atlas-Ambassadors@ripe.net https://lists.ripe.net/mailman/listinfo/ripe-atlas-ambassadors
On 2019-04-30 07:05, Evgeniy S. wrote:
Dear people from RIPE who are involved in Atlas project,
do you have any past experience to get people (who have abandoned probes) notified over email with nice message asking to connect probes back (or return to ambassador for re-distribution)?
-- Evgeniy
Hi, Yes, we're periodically reaching out to hosts of abandoned probes. We had an "all abandoned" campaign before, ant the latest one was, if I recall correctly, for Russia specifically. All of these have some effect -- there's always *some* increase in available probes. Regarding this: On 2019-04-30 02:10, Amreesh Phokeer wrote:> This is what happens when probes are distributed in the wild. In
Africa, we have the same problem, now our strategy is to give it to interested collaborators only.
Indeed, it is always much more successful to find interested people (who can at least imagine how this will be helpful for *them*) than to convince someone to sign up and "sell" the idea. I find this to be the biggest value for the ambassador program: to know the local needs and target the most responsive users. Regards, Robert
Amreesh Phokeer writes:
Hi Martin,
On 30 Apr 2019, at 00:39, Martin Levy <martin@cloudflare.com> wrote:
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online?
This is what happens when probes are distributed in the wild. In Africa, we have the same problem, now our strategy is to give it to interested collaborators only.
Yeah, that is my experience as well and why I have given up bringing probes to ICANN meetings etc. I'm getting a tad tired of these people who swear on the grave of their favorite cat that they would install them, they actually never do and never aswer mail either. This on top of those people who initial instll them and themn abandone them after some time. There is an ICANN coming up and I have still some probes to hand out but wonder whether I should bring any of them with me. jaap
Same here. E-mail is not a compelling leverage over the hosts. They must be accessible in person or the probes will be gone sooner or later. Best thing to do is to distribute only among reachable enthusiasts,and collect as soon as the flame is about to fade. By the way, has anyone noticed the drastic increase in the traffic consumed by the probes recently? Isn't it another cause for that abadonment? Regards, M. Tajbakhsh Jaap Akkerhuis wrote:
Amreesh Phokeer writes:
Hi Martin,
On 30 Apr 2019, at 00:39, Martin Levy " target="_blank"><martin@cloudflare.com> wrote:
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online?
This is what happens when probes are distributed in the wild. In Africa, we have the same problem, now our strategy is to give it to interested collaborators only.
Yeah, that is my experience as well and why I have given up bringing probes to ICANN meetings etc. I'm getting a tad tired of these people who swear on the grave of their favorite cat that they would install them, they actually never do and never aswer mail either. This on top of those people who initial instll them and themn abandone them after some time.
On 2019/05/04 14:18 , M. Tajbakhsh wrote:
By the way, has anyone noticed the drastic increase in the traffic consumed by the probes recently? Isn't it another cause for that abadonment?
Do you have example of probes where that happened? On the main probe page there is a setting to limit the bandwidth that can be used by the probe. Click the edit button next to 'General Information'. The mechanism is not perfect, but in areas where people are sensitive to bandwidth use, it may be worth making probe hosts aware of this option. Philip
Dear Philip Philip Homburg wrote:
On 2019/05/04 14:18 , M. Tajbakhsh wrote:
By the way, has anyone noticed the drastic increase in the traffic consumed by the probes recently? Isn't it another cause for that abadonment?
Do you have example of probes where that happened? This is not about one or two of the probes. Dealing with the first pack, I measured the traffic and it was not more than 200 MB per month for each probe, which was quite negligible. Now, it is nearly ~1.6 GB/month, probably following a drastic increase in number of UDM's to 200+ per probe.
On the main probe page there is a setting to limit the bandwidth that can be used by the probe. Click the edit button next to 'General Information'.
The mechanism is not perfect, but in areas where people are sensitive to bandwidth use, it may be worth making probe hosts aware of this option. I have tried setting that to smaller values, as well. Still, probes can consume a big bulk of bytes just by a continuous monthly traffic of 5 kbps. Not a big deal for some of hosts in practice, yet setting the bandwidth low doesn't help if some people have issues with those few gigabytes "swallowed" by these honorable guests, or if they are concerned about missing more.
Number of connected probes has not grown as expected during last few years, putting up to 4 times of measurement load on active ones. I am personally happy seeing that on a single device, but hosts may not be that happy. Regards, M.
On 2019/05/08 21:12 , M. Tajbakhsh wrote:
I have tried setting that to smaller values, as well. Still, probes can consume a big bulk of bytes just by a continuous monthly traffic of 5 kbps. Not a big deal for some of hosts in practice, yet setting the bandwidth low doesn't help if some people have issues with those few gigabytes "swallowed" by these honorable guests, or if they are concerned about missing more.
Maybe you can bring this to the main ripe-atlas mailing list? So far we assume that 5 kbps is perfectly normal for a probe. And all probes get a collection of measurements that results in that load. If that load is a significant issue in some countries then it is worth discussing how we should continue. Philip
Philip Homburg wrote:
Maybe you can bring this to the main ripe-atlas mailing list? I believe I did, awaiting feedback from other folks here. So far we assume that 5 kbps is perfectly normal for a probe. And all probes get a collection of measurements that results in that load. If that load is a significant issue in some countries then it is worth discussing how we should continue. Not just some countries, I may rephrase it to "some operators". Billing is different from ISP to ISP. Depending on the technology of service, their contracts may be based on the speed or be bound to the volume.
Just wondering... On 29/04/2019 22:39, Martin Levy wrote:
Hi all,
I have a basic question. If I look at the latest stats for Saudi Arabia and I see:
https://atlas.ripe.net/results/maps/network-coverage/?filter=Saudi+Arabia+(s...)
Connected: 8 Disconnected: 4 Abandoned: 43
...is there a way to find out how many of the 'abandoned' ones are actually V1 probes? A while ago, quite a few have been killed by an automatic firmware upgrade that went astray. Those probes never came back on-line and - as much as I know - cannot be revived in the field :-( FWIW, Wilfried
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online?
I could easily pick another country; however, I waned to do a measurement in KSA today, hence this is on my mind.
Yours,
Martin
Martin J. Levy <martin@cloudflare.com <mailto:martin@cloudflare.com>> Distinguished Engineer @ Cloudflare, Inc. +1 408 499 3801 (mobile) +44 7852 156845 (when in UK) https://cloudflare.com/ (web)
On 2019-04-30 10:58, Wilfried Wöber wrote:
Just wondering...
On 29/04/2019 22:39, Martin Levy wrote:
Hi all,
I have a basic question. If I look at the latest stats for Saudi Arabia and I see:
https://atlas.ripe.net/results/maps/network-coverage/?filter=Saudi+Arabia+(s...)
Connected: 8 Disconnected: 4 Abandoned: 43
...is there a way to find out how many of the 'abandoned' ones are actually V1 probes?
Yes there is. The probe archive + jq are quite powerful for answering such questions: robertk@BadGastein:~$ curl https://ftp.ripe.net/ripe/atlas/probes/archive/meta-latest > probes.bz2 % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 1070k 100 1070k 0 0 713k 0 0:00:01 0:00:01 --:--:-- 713k robertk@BadGastein:~$ bzcat probes.bz2 | jq -c '.objects[] | select(.country_code=="SA") | select(.status==3) ' | wc -l 43 robertk@BadGastein:~$ bzcat probes.bz2 | jq -c '.objects[] | select(.country_code=="SA") | select(.status==3) | select(.id<5000) ' | wc -l 11 robertk@BadGastein:~$ bzcat probes.bz2 | jq -c '.objects[] | select(.country_code=="SA") | select(.status==3) | select(.id>5000) ' | wc -l 32
A while ago, quite a few have been killed by an automatic firmware upgrade that went astray. Those probes never came back on-line and - as much as I know - cannot be revived in the field :-(
It seems about 1/4 of these are the old v1-v2 probes. Regards, Robert
FWIW, Wilfried
How can we fix the "abandoned" part? What part can an ambassador play in order to get abandoned probes back online?
I could easily pick another country; however, I waned to do a measurement in KSA today, hence this is on my mind.
Yours,
Martin
Martin J. Levy <martin@cloudflare.com <mailto:martin@cloudflare.com>> Distinguished Engineer @ Cloudflare, Inc. +1 408 499 3801 (mobile) +44 7852 156845 (when in UK) https://cloudflare.com/ (web)
_______________________________________________ RIPE-Atlas-Ambassadors mailing list RIPE-Atlas-Ambassadors@ripe.net https://lists.ripe.net/mailman/listinfo/ripe-atlas-ambassadors
participants (10)
-
Amreesh Phokeer -
Barry Raveendran Greene -
Chafic Chaya -
Evgeniy S. -
Jaap Akkerhuis -
M. Tajbakhsh -
Martin Levy -
Philip Homburg -
Robert Kisteleki -
Wilfried Wöber