New Node Provider Proposals

Thanks for the detailed explanation @SvenF, it’s much appreciated. I have some questions if that’s okay.

As you mentioned, there was an example of a single node swap proposed around the same time → Proposal: 131703 - ICP Dashboard (internetcomputer.org). Given that this can occur by shear chance (as you explained) I take it this isn’t a strong requirement (to swap a pair). Presumably a proposer should have the freedom to remove one of the pair from the proposal (if it doesn’t make sense), or in practice do they need to stick with what the tool says?

Presumably the proposer could take it upon themself to modify the proposal text? I wonder if it would be even better to state something like “optimizing for a set of decentralisation coefficients referenced by this Motion proposal → Proposal: 125549 - ICP Dashboard (internetcomputer.org)”. These coefficients don’t appear to include some important information, such as geographical distance and diversity of continents (and/or overarching jurisdictions). It seems it would be possible to optimise for the existing coefficients while actually achieving a worse decentralisation outcome in real terms. I think Proposal: 131704 is an example of this.

Assuming that Proposal: 131704 didn’t feature the MU1 data center issue, can I ask if you think it would have been a good proposal to accept (under the terms of “optimizing decentralization”)?

Hi @Lorimer if the proposal would not have included the MU1 nodes, I think yes, the proposal would be okay if it stated “optimizing decentralization” instead of “improving decentralization”. Then again, I will talk to the team and see how the summary descriptions can be improved. Adding new decentralization coefficients such as a political region and taking these into account when selecting node machines for a subnet will of course require some more discussion and time to implement.

2 Likes

Hi Marc,

Great work on getting the new nodes onboarded! I’m just going over the proposal details as part of the CodeGov team’s review process and have a question about the node machine validation documents.

The hash you’ve given doesn’t seem to correspond to the current document in the Wiki. Instead, it appears to match the earlier (15 Dec 2023) document on this page.

In the current document, you’ve given an initial Objective Value of 20, and shown that adding the 4 nodes at data centre Paris 1 would lower the Objective Value to 17, and adding 4 nodes at data centre Barcelona 1 would lower the Objective Value to 16, matching the text shown on pages 3, 5, 8, 12, 14 and 17 of the document, and thereby improving network decentralisation in both cases.

In the earlier document, you’ve given an initial Objective Value of 13, and stated that adding 3 nodes at data centre Paris 1 would lower it to 10, and adding 4 nodes at data centre Barcelona 1 would lower it to 6, however the same values of 17 and 16 as per the current document are shown in the graphical output in this document.

My guess is that the current document is probably the one for which the hash should have been given, as opposed to the earlier document. Would this interpretation be correct? My question is probably academic as the node allowances have already been approved in last year’s proposal, the current Wiki document supports this, and I see no reason to vote against the new proposal, but I’d like to check if I’ve misunderstood or overlooked anything.

4 Likes

Hey Tim, thanks for your diligent review! Yes, I believe you’re interpreting everything correctly, and apologies if theres some inconsistencies with the hash.

If I remember correctly, I experienced some troubles uploading documents to my Wiki around the time that the Topology Optimizer was introduced.

You can see comments between @SvenF and myself about this around Dec 2023 here and here.

Please let me know if you have any questions or need anything from me.

3 Likes

Thanks Marc. That all sounds fine and clears things up. It also helps to reassure me that I’m following all this properly, so that’s much appreciated too!

4 Likes

Thanks @SvenF, what if there were a competing proposal submitted shortly after by a member of the community that produces a clearly more robust decentralisation outcome (satisfying the current coeffcients, as well as addtional ones that are intuitively important). Would DFINITY take kindly to such a proposal (rejecting the original proposal in favour of the community-member-proposed one), or would this be discouraged?

2 Likes

Hi @Lorimer that’s for the community to decide, and not for the foundation. If there is a proposal that has valid arguments to further optimize decentralization, than that would definitely be something that the community would be interested in.

3 Likes

Hi Marc,

Congratulations on getting this proposal passed! Just for your info, we’ve posted some reviews of this proposal (along with others for Node Admin and Subnet Management) in CodeGov’s OpenChat channel here.

All the best,

Tim

3 Likes

Hi @SvenF . Thanks for taking on board the feedback about adding more detail to subnet management / node replacement proposals. I see that the latest proposal (131789) has a much more thorough explanation about why the change has been proposed.

I’m not sure how to verify the changes however. I’ve installed the dre cli tool just from the instructions here, but what commands or other steps should be used to replicate the result shown in the proposal?

1 Like

I agree! @sat did a great job of turning this around so quickly!

I’m not the most familiar with the dre tool, but there’s an API that can be queried for the relevant information (raw data that requires processing to pick out the necessary info).

Here’s some proposal details that I’ve put together → Subnet Management - 4zbus (Application) - Developers - Internet Computer Developer Forum (dfinity.org)

5 Likes

Thanks @Lorimer and thanks @sat! I agree with the questions you’ve asked in your other post and hopefully the team can give us some further details on these.

I’ve also voted to adopt the proposal and have posted a review with the reasoning explained in CodeGov’s OpenChat channel here.

3 Likes

@timk11 After this PR is merged, it should be possible to run:

dre proposals analyze <open-subnet-membership-change-proposal-number>

And this should re-do the decentralization analysis. There is still some polishing that needs to be done, and the documentation that needs to be written. PRs are welcome! :smiley:

5 Likes

Hi Everyone,
As some of you might be aware, we are in the process of changing our IPv6 subnet due to issues with the Geo IP database. The ISP has assigned us IP addresses that are not recognized as local, even though the latency remains the same. To address this, we’ve taken steps to acquire our own local IPv4 and IPv6 subnets. As part of this effort, we are redeploying all CM1 DC nodes from scratch. We’ve successfully migrated two nodes that were not previously in the subnet to our new IP space. Now, we have created two NNS proposals to remove the nodes that are still in the subnet and need to be migrated:

Proposal 131977
https://dashboard.internetcomputer.org/proposal/131977
Proposal 132102
https://dashboard.internetcomputer.org/proposal/132102

We would greatly appreciate your support in voting on these proposals to expedite the migration process. Thank you!

3 Likes

Hey @MalithHatananchchige. Is there a reason you posted this information after the proposals were live instead of before the proposals were live? Now that there are independent reviewers for this proposal topic, it seems desirable to improve the communications on the reasons why these changes are made. What do you think would be the best way to make the change justification available to proposal reviewers?

3 Likes

The main communication happened in the Elements Matrix Channel for NPs.
About a few weeks back before we started.

And clearly, there is room for improvement in communication. Forum would be a good idea always then

2 Likes

Actually, this would be good if there was a different topic on the forum to keep the reviewers in the loop on ongoing node maintenance, as there are many reasons for redeployment. Like hardware failure, ISP issues, or switching to a different ISP, Purchasing your own IP space

What do you think?

3 Likes

Oh that’s very interesting. I’m not familiar with that channel. Is it open to the community and/or reviewers? Would it be possible for the CodeGov team (@ZackDS @timk11 @LaCosta @wpb @Zane @cyberowl) as well as @lorimer to get access?

Yes, I think the forum is likely the best place for this communication. I believe it has already been suggested to @lara and @cryptoschindler that a tag for each proposal topic should be created, but separate forum topics could also be used. Do you think all node providers would be willing to follow a work process that includes adding more information about why changes are proposed? It seems likely they would be willing, especially if everyone knows that there are independent reviewers in addition to DFINITY who are reviewing the proposals.

2 Likes

Yes the NP matrix channel is public and anyone can join
https://wiki.internetcomputer.org/wiki/Node_Provider_Matrix_channel

Even though I can not speak on behalf of all NPs I think they would and I definitely would not mind sharing why a proposal on Node management would happen with the community.
This way we can bridge the gap on any communication issues. Also, I intend to update the wiki on my learnings.(Anyone can update it)

https://wiki.internetcomputer.org/wiki/Node_Provider_Documentation

3 Likes

Thanks Malith for this info! I’ve gone on to the NP matrix channel and joined up. The information there certainly gives a clearer context to these two proposals. If NPs can post something on this forum for these sorts of proposals it would be a big help as it’s easy to access without having to go through sign-up steps.

Personally I’m in favour of a forum post link being included within any technical NNS proposals, wherever in the forum the post might happen to be. This covers a few other proposal types as well so it might warrant a separate discussion in this forum.

4 Likes

Great, I don’t think any NP knows that subnet management is under independent review and should be posted on the forum. I don’t mind posting on the forum—the idea that @Lorimer brought up seems to work well. You might need to inform all NPs or encourage this practice. In the meantime, as mentioned, I will update the Wiki to include the correct process for updating subnet forum posts when intending to remove a node for maintenance.

5 Likes