RIPE 79

Date: 17 October, 11:00 - 12:30
Chairs: Ignas Bagdonas (Absent)
Interim Co-Chairs: Rob Evans, Job Snijders
Scribe: Adonis Stergiopoulos
Status: Final 

0. Administrivia – WG Chairs 

The presentation is available online:
https://ripe79.ripe.net/presentations/99-rtg-wg-agenda.pdf

The Chairs announced that the minutes of the previous meeting at RIPE 78 have not been sent to the mailing list yet. They will be finalised in the next couple of weeks and will be shared with the community via the mailing list. If there are no objections, the minutes will be approved.

Hans Petter Holen, RIPE Chair, sent a request to the RIPE Routing Working Group (WG) mailing list asking for volunteers who would like to become the next WG Chairpersons. Paul Hoogsteder and Job Snijders declared their candidacy. Rob Evans asked the attendees if they agree to appoint Paul and Job as Co-Chairs. There were no objections, therefore, Paul and Job are Co-Chairs of this WG.

Rob congratulated the new Co-chairs and made an observation about the WG Chairpersons selection process. It feels the process can be quite intimidating for newcomers trying to volunteer to become Co-Chairs and running against candidates with a lot of support from the mailing lists. Rob suggested that future candidates first make themselves known to the existing Co-Chairs. Afterwards, their names will be simultaneously announced to the WG.

1. RIS BGP Route Collection – Florian Obser and Emile Aben, RIPE NCC

The presentations are available online:
https://ripe79.ripe.net/presentations/94-RIPE79_RIS_Update_BD.pdf
https://ripe79.ripe.net/presentations/105-ripe79.ris-dev.emileaben.reviewed.pdf

Emile shared some ideas on how to further develop RIS and asked attendees for their input.

Blake Willis, Zayo / iBrowse, commented that RIS is super useful. Blake asked if they take a feed from the route servers at the IXs as well. Florian replied that they currently do not as they do not know how to represent this data.

Martin Levy, Cloudflare, said that RIS has become the “de facto” source of data and they should be very proud of that. Martin commented that currently they have a very passive role, asking people if they will feed, and asked the presenters if they would consider taking a more active role finding their collection points. Emile agreed.

Andrei Robachevsky, Internet Society, pointed out that in the presentation there are multiple router collection efforts mentioned and not only RIS. Andrei asked the presenters if they have analysed diversity for RIS views or PCH or collections and if there is any mileage of combining those efforts to address this issue. Emile replied that there is an effort on the way and recommended watching a RACI talk from Franziska Lichtblau who presented on this topic earlier that week.

Paolo Lucente, NTT, suggested they start looking into BMP. Currently, a lot of work is taking place at the Global Routing Operations (grow) at IETF – they are adding TLVs and will transport BGP information and any metadata. Paolo suggested to the presenters that if they want to collect more information, they need to use BMP all the way.

Sebastian Castro, InternetNZ, agreed with Martin’s earlier suggestion and proposed nudging people to having a collector rather than forcing it.

There were no other questions.

2. BGPalerter – Massimo Candela, NTT

The presentation is available online:
https://ripe79.ripe.net/presentations/111-BGPalerter_ripe79.pdf

Michiel Klaver, a remote participant, asked if they would consider offering BGPalerter as a hosted solution for others. Massimo and Job replied this is something for them to consider and will need to ask their manager approval to spend time on that.

Ivan Beveridge, IG, asked if there is a configurable way to send messages to a locally hosted equivalent to Slack, such as Mattermost. Massimo replied that the reporter does that, essentially uses IX queries or web book. If Mattermost supports that, you essentially have to change the code, change the JSON format and the end point, and create your own reporter.

Will van Gulik, Nimag Networks, commented that installing the BGPalerter was very easy and also agrees that a hosted service would be great. Massimo thanked Will and said that the query they do to RIS is really lightweight both in terms of CPU and memory. You will only get the full stream if you have done a very complex query, otherwise, it uses the bare minimum.

Robert Kisteleki, RIPE NCC, said this work is excellent. The RIPE NCC is currently working towards moving RIS Live into production. Once that happens, they will be interested in working with NTT and intergrating BGPalerter into their system. Robert also commented that it is entirely possible for the RIPE NCC to take up such a hosted service and asked the community to let them know should they think it would be useful.

Randy Bush, Arrcus & IIJ, agreed with Will that this is very good work. Randy asked if it puts a wider load on the RIPE NCC’s servers if we aggregate users and what is the trade-off space. Massimo said they tried to optimise it and the subscriptions are going to be minimum - less than a megabit even if you get the entire stream. Randy replied that it is the backend to RIS that he worries about. Massimo answered that they have ran tests since his time working at the RIPE NCC and it seems to be holding on for now. NTT will wait to see what happens if many people use it. Job added that the tool was specifically designed under the assumption that the RIPE NCC's RIS Live API would not be the only thing that feeds it but that other things can feed into it as well.

There were no further questions.

3. Usage of Contemporaneous Border Links (CBLs) in Internet Multipath Routing – Jie Li, UCL/RACI

The presentation is available online:
https://ripe79.ripe.net/presentations/55-RIPE79-GBLUsage.pdf

Job asked Jie to go back a few slides (Slide 16) to the listing of the AS Numbers. Job then asked why NTT is not in this list. Jie replied that these ASNs are only involved in the cases and it does not mean they are not included in the measurements. Job thanked Jie for clarifying.

Ivan Beveridge, IG, commented that it would be useful to have some feedback from the carries to understand why Jie was actually looking into these statistics.

There were no further questions.

4. RPKI Resilience, How Trustworthy is our Trust Anchor? – Nathalie Trenaman, RIPE NCC

The presentation is available online:
https://ripe79.ripe.net/presentations/96-RPKI-Resilience.pdf

Randy Bush, Arrcus & IIJ, thanked the RIPE NCC for diligence. Randy wanted to reassure people that even if there is an outage, there is no need for panic. Nathalie agreed that if RPKI goes down, everything will revert back to “Unknown”. This might not be great but at least it is not “Invalid”. There is a big difference there. Randy added that Validators will use stale data if they cannot get frash ones. Nathalie agreed and thanked Randy. Job also thanked the RIPE NCC’s efforts to ensure that the RPKI operational service is of the highest quality.

Ruediger Volk, Deutsche Telekom Technik GmbH, said there is some way to go until we fully understand all the parts that are needed. Ruediger commented that a review by a third party is fine but not sufficient as the published details should be formally disclosed in the RIPE NCC’s Certification Practice Statement (CPS). The current CPS is very old and cannot provide this information. Nathalie agreed and replied that they are currently in the process of fully reviewing that.

Peter Hessler, KLEO Connect GmbH, asked about the downtime outages and what sort of server redundancy, and especially geographical redundancy, the RIPE NCC currently has and plans to have in the future. Nathalie replied that they have the repository with 100 percent uptime. This is due to the RPKI Repository Delta Protocol (RRDP) running on Amazon clusters and rsync running on local servers. Looking into the core infrastructure, two of their five servers were replaced by virtual machines. The other downtime was due to the LIR portal outage. Nathalie commented that you cannot always keep everything running.

There were no other questions.

5. Validating MANRS of a Network - Andrei Robachevsky, ISOC / MANRS

The presentation is available online:
https://ripe79.ripe.net/presentations/93-201910-Validating-MANRS-RIPE79.pdf

Peter Hessler, KLEO Connect GmbH, asked if MANRS is primarily analysing the configurations of the router or sending live traffic. Andrei declined and replied that the input is your configuration and some metadata. The output is how this configuration matches the requirements. Peter responded that this sounds a very interesting project, both telling MANRS that network operators are compliant and vice versa.

There were no further comments.

6. AOB

The video recording can be viewed online:
https://ripe79.ripe.net/archives/video/262/

Benno Overeinder, NLnet Labs, talked about the next generation of the WG leadership and proposed to the WG to train a young person. Job agreed and commented that they can look into teaching a young person what it means to be a Chairperson over the next two RIPE meetings. Job added that some work also needs to be done on the WG Chairpersons selection process, emphasing a degree of diversity, churn in leadership, and the notion of having a trainee.