[RIPE Atlas Ambassadors] Deploying RIPE Atlas probes in more locations and networks

Really good question..

An ISP or organization is hard pressed to capture metrics outside their system I contend — unless you pay for a commercial service and even then you don't get the span of coverage that Atlas has.

What I have done is use the probes  to be part of the overall measurements and a way to accrue credits to run tests to monitor availability reachability qualities OUTSIDE my overall network.

If your 'network' spans multiple regions/geo-locations, have a few probes and capture reachability metrics to capture quality of experience as a matter of operation.

  *   Often it's "I think it's slow, it must be the network"
  *   Or "I can't get to X but get to Y, it must be the network"

But is it really 'the network' all the time?
Highly unlikely, but when it is, how could you triangulate a problem that is intermittent by time and locale?
If you have measurement nodes (probes), this becomes slightly easier and is one datapoint to add to the diagnosis mix.
They aren't a replacement for Nagios/Big Brother/PerfSONAR infrastructures, but there is no way you can economically achieve the same network coverage as Atlas does.

For example, I have a service located in Canada and identified Canadian probes to attempt to do a certificate check as a way to exhibit reachability (not a content check) to my host.  This allows me to assess 'How reachable am I across Canada from these probes to my host?'.  I have done the same for 'outside the country to my host' reachability as well.

The other aspect is that probes can be operated on consumer connections. These allow a true 'end to end' measurement that would more accurately depict user experience — not only from Internet exchange(IX) to IX.  This is sometimes overlooked.

To sum it up:

  *   The probes add additional capabilities for network measurement and diagnosis of problems.
     *   understand your use cases & have a metrics plan to begin with.
  *   They can run autonomously without you setting up infrastructure  regardless of IX or home cable modem install site.
     *   Slim to small operational management overhead.
  *   They allow you to tap into a global measurement network without having to run the backend stats aggregation at scale.
     *    a very big plus!

Chris.
___________________________________________________________________________________________
Chris Phillips
Technical Architect, Canadian Access Federation | CANARIE| chris.phillips at canarie.ca<mailto:chris.phillips at canarie.ca> | W: 613.943.5370 |GPG: 0x0380811D


From: RIPE-Atlas-Ambassadors <ripe-atlas-ambassadors-bounces at ripe.net<mailto:ripe-atlas-ambassadors-bounces at ripe.net>> on behalf of "dp at datasoftcomnet.com<mailto:dp at datasoftcomnet.com>" <dp at datasoftcomnet.com<mailto:dp at datasoftcomnet.com>>
Date: Tuesday, August 25, 2015 at 12:18 AM
To: "emile.aben at ripe.net<mailto:emile.aben at ripe.net>" <emile.aben at ripe.net<mailto:emile.aben at ripe.net>>, "sebastian at nzrs.net.nz<mailto:sebastian at nzrs.net.nz>" <sebastian at nzrs.net.nz<mailto:sebastian at nzrs.net.nz>>
Cc: "ripe-atlas-ambassadors at ripe.net<mailto:ripe-atlas-ambassadors at ripe.net>" <ripe-atlas-ambassadors at ripe.net<mailto:ripe-atlas-ambassadors at ripe.net>>
Subject: Re: [RIPE Atlas Ambassadors] Deploying RIPE Atlas probes in more locations and networks

Hello all,
Can someone share an elevator pitch on why an ISP or an organisation that is on 24/7 needs to put a probe in their  network.
What are the benefits they would get?
Has someone prepared a case study.

Regards DP
9849111010.

Sent from my phone. Please excuse brevity and typos if any.


-------- Original message --------
From: Emile Aben <emile.aben at ripe.net<mailto:emile.aben at ripe.net>>
Date: 25/08/2015 09:25 (GMT+05:30)
To: Sebastian Castro <sebastian at nzrs.net.nz<mailto:sebastian at nzrs.net.nz>>
Cc: ripe-atlas-ambassadors at ripe.net<mailto:ripe-atlas-ambassadors at ripe.net>
Subject: Re: [RIPE Atlas Ambassadors] Deploying RIPE Atlas probes in more locations and networks

On 24/08/15 23:53, Sebastian Castro wrote:
>
>
> On 24/08/15 11:27 pm, Vesna Manojlovic wrote:
>> Dear ambassadors,
>
> Hi Vesna:
>
>>
>> You’ve helped us do a great job in distributing RIPE Atlas probes to
>> many networks and regions of the world, but we still have work to do!
>> There are a few ways to find out exactly where we need your help:
>>
>> 1) Networks with the most users and fewest probes
>>
>> In order to increase topological diversity, it would be good to "target"
>> the users of the networks that don’t have a lot - or any - coverage.
>>
>> This RIPE Labs article lists the top 20 ASNs:
>> https://labs.ripe.net/Members/emileaben/improving-ripe-atlas-coverage-what-networks-are-missing
>>
>>
>> Or you can see the raw-text table here, sorted by the network size:
>> http://sg-pub.ripe.net/emile/atlas_eyeball_coverage.txt
>>
>
> Emile's work is really valuable and much appreciated. Is it possible to
> have a date of generation on that file? It seems to be the same as few
> months ago, considering we have deployed probes in the last two months
> that increase coverage.

This was a one-off, timestamp on the file says I did that on 26 June.
I've just put a new one here:
http://sg-pub.ripe.net/emile/atlas-coverage/

>> 2) Comparing population density with the distribution of RIPE Atlas probes
>>
>> Emile Aben used CartoDB to plot, population and probe density on the
>> same map:
>> https://emileaben.cartodb.com/viz/3caa5144-823e-11e4-8025-0e018d66dc29/public_map
>>
>>
>> https://labs.ripe.net/Members/emileaben/distribution-of-ripe-atlas-probes
>>
>> 3) Comparing coverage between countries
>>
>> This map shows the percentage of probes per country:
>> https://atlas.ripe.net/results/maps/density/
>>
>> Whichever method you choose, we trust your instincts and your knowledge
>> of the local community to make the right decision and give probes to the
>> best people!
>
> We have a different method, that uses address space to determine which
> ASNs are relevant to cover.
>
> The eyeball method works for ISP that serve customers, but not for those
> that provide transit. Our methodology, that picks a BGP table dump and
> calculates the ratio of /24 versus number of probes helps us to have
> more coverage and to quickly identify targets.

Heh. We had a discussion last week in the office which boiled down to
pretty much what you describe here, as an alternative to 'eyeball' data.
I expect more in this space soon :)

Cheers and many thanks to all our ambassadors for the great work they
are doing!
Emile

_______________________________________________
RIPE-Atlas-Ambassadors mailing list
RIPE-Atlas-Ambassadors at ripe.net<mailto:RIPE-Atlas-Ambassadors at ripe.net>
https://www.ripe.net/mailman/listinfo/ripe-atlas-ambassadors
-------------- next part --------------
An HTML attachment was scrubbed...
URL: </ripe/mail/archives/ripe-atlas-ambassadors/attachments/20150825/7c990d77/attachment.html>