Re: [mat-wg] Some suggestions for stat.ripe.net


Hi Mauricio,

Thanks for your contribution!

On 04/05/2011 14:39:08, Mauricio Vergara Ereche wrote:
> I think that this is a great project! It has a nice interface and is very 
> useful to get an "one page" info for everything. Excelent work!
> 
> I have a few sugestions for the new system that I think can improve the value 
> of this project:
> 
> * Add an input field for each box, so the owner of the resource (or maybe 
> anyone, I haven't think this too hard) can add more info about that 
> particular 
> item (ie: on geolocation, add the "true" location, on assignments within 
> allocation - well, it's obvious :-P)

This is a nice idea, we'll need to think about if/how it could be done
on a per plugin basis. If you (or anyone) have specific ideas for any
plugin, let us know.

Two (of several) considerations:

1) For most resource data, the RIPEstat output is taken from mirrored
RIPE NCC sources. We can't update the original via RIPEstat. And in most
cases, unless we can properly identify the true resource holder, we
couldn't anyway

2) In the specific case of geolocation, this data is sourced from
MaxMind, and we have no way of correcting it. We could perhaps store an
offset...

I realise you're probably not a RIPE NCC region resource holder, but
those who are have the ability to check and edit location data for their
own resources via our LIR Locator. Perhaps one day this data set will
feed into the RIPEstat geolocation module:

http://labs.ripe.net/Members/kranjbar/ripe-ncc-lir-locator/

> * On the visibility test, when you get a "!" and a porcentage (let's say, 
> 80%) 
> it would be great to know if there was 80% of the probes over there that 
> wasn't able to "see" the resource, or if 80% of the time was not visible, or 
> if 80% of the times failed to check. Anyhow, I think it would be great to 
> know 
> more about this topic. If something failed, the hour of the fail, or the 
> route 
> to get to that resource.

The methodology text should make this clear - it's far too vague at
present. We can work on that - it's on our todo list.

What I'd like to hear from you and this list is what optional metrics
you'd like us to use to calculate this figure? What is meaningful and
useful to you? We can support multiple options, what would you find most
interesting/useful?

> * Can you add a new box for do some kind of lookinglass over the resource 
> from 
> different locations/peers? that would be awesome! When i have to test some 
> anycast cloud, i have to fly all over the looking glass that i can find, and 
> then do my own analysis.

If you join the RIPEstat demo on Tusday
https://labs.ripe.net/Members/markd/ripestat-live-demo-5 and you might
be in for a nice surprise :)

All the best,

Mark

-- 
Mark Dranse
RIPE NCC