[Atlas-anchors-pilot] Q&A, part 1

[Vesna Manojlovic]

Dear colleagues,

thank you for your interest in the Atlas Anchors pilot , and thanks for 
your questions till now! Here is an attempt to give some answers.

(housekeeping tip: since there a lot of questions,
*Please open a new thread when you want to discuss a specific topic* )

Basic concepts were given in this article:
https://labs.ripe.net/Members/dfk/ripe-atlas-anchors

Q&A:
====

Q: Is it possible to use current TTM hw?
A: No.

Q: I am not coming to the RIPE meeting. Can I take part in the 
discussion in some other way?
A: No problem, there is a discussion list:
     https://www.ripe.net/mailman/listinfo/atlas-anchors-pilot

We will be talking about this at mat-wg list & ncc-services-wg.
Here is the meeting plan:
https://ripe65.ripe.net/programme/meeting-plan/
There will be possibility for remote participation:
http://www.ripe.net/ripe/meetings/ripe-meetings/take-part-remotely
More details will be available during the meeting, on ripe65.ripe.net

Q: I'm not a member of ripe ncc. what are my options?
A: The only option is to become a RIPE NCC member:
how-to: https://www.ripe.net/lir-services/member-support/become-a-member

Note that the membership fees are under discussion currently: 2013 
charging scheme changes proposals:
https://www.ripe.net/internet-coordination/news/announcements/draft-activity-plan-budget-and-charging-scheme-options 


Q: I can't afford the HW, but i can offer bandwidth & all the other help.
A: Sorry, there is nothing we can do. We do not have the means to give 
away hardware.

Q: What are the HW specs?
A: In a separate message on the mailing list, to be able to link to it 
easily:
https://www.ripe.net/ripe/mail/archives/atlas-anchors-pilot/2012-September/000000.html

Q: still waiting for the rationale why the box would need to cost 3000 
EUR...
(I want an Atlas Anchor, focused on "target for the measurements, and 
run local tests", not an "it can run every service the NCC might come up 
with in the next 5 years" box - so no interest in a local
instance of k.root, a local RIPE whois database, a RIS collector, etc.)
A: The bulk of the cost here is operational. If we have to operate 50 or 
so anchors with the resources we have we cannot skimp on the hardware 
and then work around the limitations we incur. The hardware needs to be 
reliable with good remote management and enough headroom. Especially not 
in the pilot phase. We suggest you get one anchor in the pilot phase and 
help us gain the experience to maybe develop a less costly option. If 
you want us to take the operational responsibility, there are some 
standards we have to set.
For example, "DRAC" component adds additional costs, but is important 
for remote management.

Q: More information about your needs in colo/bandwith?
A: Colocation: The box is rack-mounted 1U, and dual power is preferred.
We expect not more then 1Mbit traffic to/from the Anchor;  we strongly 
prefer locations with  IPv6 connectivity. Anchor should not be behind 
NAT. Firewall settings: The public IP address should allow all in- & 
outbound traffic; admin interface should allow all outbound connections, 
and incoming from RIPE NCC  space.

Q: Are you planing to make jitter measurements available?
A: Not in the short term.
We would like to know:
- what is the use case for jitter measurements?
- what is the resolution/varience needed?

We are planning to add one-way packet loss measurements in the future, 
to demonstrate that there is value in having access to the target.

Q: How will the (number of probes / 200-500?) for baseline measurements 
  be chosen?
A: RIPE NCC will chose them, based on the topological closeness.
We will of course take wishes of the anchor host into consideration.

Q: Which measurements will be available towards the Anchor?
A: We are planning on the slow-start principle: ping & traceroute to 
start with, with adding DNS & SSL later, followed with HTTP when it is 
implemented for all.

Q: Which visualizations of the measurements results will be available?
A: 	* A map of ping RTTs;
	* reachability map for pings, and HTTP
	* traceroute visualization

Q: How can we get the data of the measurements?
A: usual ways: web interface, API, visualizations.

Q: How many additional credits will be AA hosts earning?
A: Initial thinking is equivalent to about 20 Atlas probes for just the 
target functionality
and maybe additional credits once the probe functionality is there. We 
are open to community discussion.

Q: What are the Terms and Conditions going to be?
A: This is also going to be finalized during the pilot phase.
We will post the draft T&C to this list before the RIPE meeting.

For the shorter read, here is a link to the message Daniel Karrenberg 
sent to MAT-wg list:
http://www.ripe.net/ripe/mail/archives/mat-wg/2012-June/000217.html

Q: Will Atlas Anchors be available for other measurements?
A:  We prefer if the AA hosts would make their AA IP address widely 
available for measurements, and advertised as such; for example, by 
registering the AA address in the "pingable:" attribute in the 
route&route6 object : http://tools.ietf.org/html/rfc5943


Suggested phases of deployment:
===========================

- September: discussing Terms & Conditions, and services details,
- October: installing first AA as targets, ping & traceroute only
- November: installing "probe SW" on AA
	 adding more AA as targets
- December: adding up to ten(*) AA this year
	adding more built-in measurements towards AA
- Q1 2013:
	- migrating DNSmon to AA
	- adding 10 (*) more AA
- Q2 2013:
	- adding 10 (*) more AA
- by the end of 2013 plan is to have about 50 Atlas Anchors


(*) Number 10 is provisional, based on the planning before the input 
from the pilot participants, and will be adjusted.

Please open a new thread when you want to discuss a specific 
topic/question!