diff --git a/sig-providers/meetings/015-2024-03-27.md b/sig-providers/meetings/015-2024-03-27.md new file mode 100644 index 00000000..50c0683c --- /dev/null +++ b/sig-providers/meetings/015-2024-03-27.md @@ -0,0 +1,188 @@ +# Akash Network - Providers Special Interest Group (SIG) - Meeting #14 + +## Agenda +- Update from Praetor on all of the work that they have been doing. +- Open Discussion on all things related to Akash Providers. +- Update from Scott on provider feature discovery +- Discussion on provider feature discovery +- Provider Audits Updates + +## Meeting Details +- Date: Wednesday, March 27, 2024 +- Time: 08:00 AM PT (Pacific Time) +- Recording: Coming Soon +- [Transcript](#transcript) + +## Participants +- Andrew Hare +- Andrey Arapov +- Anil Murty +- B S +- Damir Simpovic +- Dani +- Deathless +- Deval Patel +- Jigar Patel +- Nathan Ward +- Rodrigo Rochin +- Scott Carruthers +- Thomas Greco +- Tyler Wright + +## Meeting notes +- **Welcome and Previous Meeting Recap**: + - Recap of lively discussions from the last meeting on gas fees and provider audits. + - Update on the creation of a working group for provider audits and implementation of provider incentives. + +- **Provider Announcements**: + - **Scott Carruthers**: + - Recent announcements on updating provider attributes to include RAM and interface specifications. + - Importance of updating these attributes to ensure compatibility and accurate bid responses. + - Documentation updates to reflect these changes. + +- **Provider Attributes Updates**: + - **Scott Carruthers**: + - Providers should include RAM and interface tokens in their attributes for better bid specifications. + - Updated the bid price script to accommodate pricing differentiation based on GPU interface. + - **Damir Simpovic**: + - Suggested applying similar logic to processor, memory, and storage attributes. + - **Anil Murty**: + - Explanation of efforts to standardize provider attributes and automate their update through provider inventory service. + - Importance of displaying GPU pricing to potential customers and ongoing work to make this information available. + +- **Provider Audits**: + - **Tyler Wright**: + - Reminder for audited providers to update their contact information. + - Importance of community members spreading this information to ensure all providers are up-to-date. + +- **Praetor Team Update**: + - **Jigar Patel**: + - Ongoing work on feature discovery and addressing provider issues. + - Busy with roadmap tasks and provider support. + +- **Open Discussion**: + - **Tyler Wright**: + - Network upgrades and parameter changes positively impacting gas fees and provider operations. + - Call for community assistance in spreading the word about updates and upgrades. + +- **Closing Remarks**: + - **Tyler Wright**: + - Encouraged attendees to share provider announcements and updates with their networks. + - Mentioned upcoming meetings: Sig clients monthly meeting and steering committee meeting. + +## Action Items +- All Providers to update RAM and interface attributes in their configurations. +- All Providers to Ensure contact information is current for provider audits. +- Core Team to Continue working on GPU pricing page and automation of provider attribute updates. +- Praetor Team to continue support and updates for feature discovery. + +# **Transcript** + +_This editable transcript was computer generated and might contain errors. People can also change the text after it was created._ + +Tyler Wright: All right, welcome everybody to providers monthly meeting number 15 during the provider special interest group folks come from the community come together to talk about all things related to being a provider on We get updates from the page war team, which is a tool that many folks in the community use to build and manage their providers. We talk about any provider related documentation or announcements that have been made over at the last monthly meeting number 14. There was some Lively discussion around some gas fees related to r attributes. There was some conversation about provider audits in general and there has since been a working group created around that + +Tyler Wright: there was some discussion around the provider incentives that is since gone chain passed on chain, and there's a number of folks that have already come on board. I think a nearly even hinted at some of those things at the last six providers meeting were at a meeting since then about some of those relationships. So there's again been a lot of action over the last month in terms. So what's thing going on related to their cash provider? I know that there's been a number of announcements that have been made recently as well related to + +Tyler Wright: Related to the provider. So I know that at the part of today's agenda we could touch on any of the things I talked about the top of a call. We'll check in with the page war team. to see if they have any updates on anything they're in but I do want to start to see if Scott or Andrey want to talk about anything specific related to some of the provider announcements. They've thrown in the chat over the last couple of days. So I will hand it over to Scott Andrey to get us kicked off. We're talking about some of those announcements and then we'll go into the typical agenda. + +Scott Carruthers: Yes, I'll just briefly cover I think the last couple of announcements have been related to attribute So several days ago. We announced to providers that they should begin including. They Ram token and their provider attributes and the reason for that is so if you look at an sdl and you're wants gpus in your bid you define the model and then the deploy or else has the ability to specify a Token or a RAM amount. So the reason that we sent out an announcement recommending that all providers update their attributes would they the ram token? + +Scott Carruthers: Is before that if a deployer decided to include Ram in the sdl and the provider didn't have that attribute defined then the one embed. So we had one announcement to please update mirror. So we had a number of providers that were already using those kind of attributes, but they were typically Only providers that had GPU models that had different ramp specs. So as an example some of our providers that are sourcing a0 Nvidia a100s would have already included those type of attributes because they need to specify if the a100 had 40 or 80 gigs of memory. But then if you had a RTX 30 70 or something like that, that only had one ram footprint providers typically weren't including those attributes. + +Scott Carruthers: But again, we want all the providers to update with the ram token and attributes because if I deployer decides to include that they're deploying with an RTX 37 day, but they find some value including Ram in the dly. So that was one provider attribute update and then last night yesterday we had. a request to request how providers to update their attributes an additional iteration in this time is to include the interface attribute or token, so + +Scott Carruthers: The sdl supports the ability again. This is all GPU related. The sdl allows the ability to include interface model So currently is that's just pcie or xxm. So if a deployer decides to include the interface in their sdl again the provider Like that on that if they have a attribute that includes C interface specification. So those are just a lot there. You can follow the announcements in the Jeep in the provider announcement Channel, but that's been the most significant updates and announcements over the last weekly as all related to provider attributes and ensuring that providers bid when certain specs are included in the sdl some tangential. + +00:05:00 + +Scott Carruthers: I'm updates in those regards and yesterday or this morning included the interface for GPU specifications. if I interface is included in the bid price script now has the logic to do pricing differentiation depending on the interface. So again, you could have a 100 and it could have a pcie bus or interface or I could have a xxm and we should have the ability to do pricing differentiation if they Tenant, or the deployer includes the interface and the sdl so not only do we have the announcement out to Providers to update their attributes, but bid price script has also been updated to again do pricing differentiation based on the interface of the GPU that's included in the STL. so off the top of my head. Those are the most significant announcements also. + +Scott Carruthers: everything that I just talked about from a sdl perspective has been updated and documentation. So we've also had a number of Provider documentation. updates see how the top of my head. Those are the most significant updates. + +Tyler Wright: Thank you very much, Scott. + +Scott Carruthers: does anybody any other providers in the audience have any questions and those regards? + +Scott Carruthers: Okay. + +Tyler Wright: I had damir. + +Damir Simpovic: Hello, everyone. I do have + +Damir Simpovic: I don't know. how important it is to have it implemented right now, but + +Damir Simpovic: I think the same logic would be. really prudent to apply to processor memory + +Damir Simpovic: storage, Etc. So, yeah. just for me. Yeah. + +Scott Carruthers: Yeah, yeah Yeah, so I think there's gonna be a lot of announcements and updates that are Generally in the umbrella feature Discovery. So the other thing. that is very top of mind right now as we want to take out How much manual actions are necessary in this process so feature brought us light years ahead and that you don't have to manually label your nodes any longer for gpus or other We no longer have to specify what type of menuing label nodes for persistent storage type in GPU types, and it's certainly Our intention that we're going to bring those feature discovery. + +Scott Carruthers: Aspects into attributes as So providers won't have to manually Define these attributes. So we're going to be using feature Discovery in the near future to update that and then simply to your point Not only would that be applicable to gpus but it would also be applicable to CPU and memory and so yeah, I think there's going to be a significant amount of feature Discovery related introductions in the outcome in few months + +Scott Carruthers: so that you had your hand raised did you have + +Anil Murty: I think I was going to say some of the things that you said already So yeah, as wrest Scott said basically we've had discussions about standardizing on provider attributes for a really long time. I think I may have included you on some of those discussions as well. There was a document that we started working on a while back trying to collect all the possible attributes that we could think of and come up with a standard naming convention for them as well as a structure in terms of where they sit in the IRT. So we started thinking about that a while back. We never got to actually + +Anil Murty: Implementing and enforcing that across the providers and as we were trying as we were working through and thinking on how we want to do that the provider inventory service came about and so the thinking was rather than having the updated. It'll be really nice if the inventory service could read. the capabilities on the provider automatically and updated for the providers rather than relying on people having to do it manually. So that's basically what the story of this whole thing is and so the reason why this kind of became a little bit of an urgent request this week was Driven privately by as needing to us wanting to come up with a pricing page for gpus. So one of the pieces of feedback we have gotten from a lot of customers in the last few weeks is it's great that you guys have a lot of gpus and when we get on conversations with these customers we show them. Hey robot all these gpus you can know submit a bit request you can get + +00:10:00 + +Anil Murty: a bunch of bids from whole bunch of providers on the network and they're very impressed by all that but the problem is that if somebody comes to the web page, there is no way of knowing that we have all these gpus on the network and they're available for such amazing prices, So what we realized is that we need some way of displaying what the prices specifically of the gpus but in general compute resources on the network are for anybody that comes to the network and typically anybody that comes to a web page, so It's kind of a little bit of an image not emergency. But a kind of an urgent request that we're handling one off just to deal with being able to surface up pricing on gpus and then when you talk about pricing of gpus + +Anil Murty: The type of interface that the cluster has typically PCI versus S6 and 5 or 6 and 4 that makes a significant difference on the price that the provider charges. So that's where these requests are coming in so just to provide a little more context and why number one we decided to handle just these two attributes one off And what's kind of the overall plan going forward, hopefully. Thanks. + +Tyler Wright: I had the ball. + +Deval Patel: Then that would be good for us because then from the provider side we do manage pricing at this point of time, but we do it hard way we store those in database and things and if they do it, let's say if they update their pricing directly on the price script and we don't have track of that. So that's always one issue from our side as well. + +Anil Murty: Yeah, so I think we'll have to check with the Maxi on whether he's able to expose it as an API. He's actually working On that API right now whether it becomes a public API, I think most likely it will be either way. We'll figure out a way to get it to you guys. But one of the challenges essentially just to go a little bit deeper on that is that there is no way today to extract specifically the GPU pricing from a provider directly because number one we don't actually get pricing from the pricing script directly. We only find it out when the provider responds to a big request and when the provider responds to a better request, there's quantifically to the Entire CPU of this memory storage. so max C has been working on a workaround to that he's trying to figure out the lowest. + +Anil Murty: bid request that has gone in with mostly a GPU request and then use the responses to that to figure out what the price of GPS maybe so it's not like a super accurate solution right now and we have improve it but that's kind of where + +Tyler Wright: Excellent. Does anyone have any other questions? Related to what Scott was just talking about. + +Tyler Wright: Cool kind of a natural segue before we again talk again update from the pray to our team. I know that Andy Andrew Andre, excuse me, had put a message in provider announcements about r attributes and excuse me about provider Audits and asking folks to update at least email or an organization if they have So if you have seen that in your audited provider, please look to update or if you're an Insider or somebody's active in the community, please let folks know again. We just want to make sure for things like this. There's a number of folks that go through the upgrades may not be as active and Discord, so they're hard to get a hold of and they may not know what's going on. And so we want to make sure that again folks that are in the community on this call or + +00:15:00 + +Tyler Wright: any other providers that they know or see around Discord or elsewhere know about the most recent information related to upgrades so that again all providers are as up to date as possible and then making sure that there is at least a contact information for audited providers. So that's been added recently to documentation as well as to the provider audit process moving forward Andres or anything specific you want to talk about that or + +Andrey Arapov: I guess nothing from my end. + +Tyler Wright: like other RF + +Andrey Arapov: I put it in announcement. I think it's quite clear there and people already started deleting their provider attributes. So if anyone has any additional questions, they can find me on this card or just Drop messaging providers Channel tag me or some other core member of the team. + +Tyler Wright: Thank you. All right, continuing to move along with the agenda. And again, if anybody has anything they want to talk about feel free to drop it in the chat or we'll talk to the prey to our team real fast. And then feel free to raise your hand after that. But I do want to hand it over to the Creator team. They usually give us an update on a monthly basis on things that they've been working on or just things that they're seeing related to a cost provider so printer team take it away. + +Jigar Patel: Hey guys, good morning. So from the last month as we discussed in last meeting, we are updating all the latest provider update that has the feature Discovery portion it So we gave that update with the recent Sergeant providers. We are very very busy with all the providers solving their issues plus working towards our roadmap as well. So yeah, that's a short update from us and I think so at this point of time, but we'll continue work on it. Thank you. + +Tyler Wright: Thank Again presentation does a great job of answering questions? They see inside the ecosystem pray to our Channel inside of Discord as well as in other channels. So I know that the crater team is very active all over Discord. So Much appreciated feel free to reach out to them or anybody from the core team if you have any questions there. + +Tyler Wright: Again out of the last meeting. We talked a little bit about some of the gas fees some of the things related to provider. I think it was a great Lively discussion on that. There was a network upgrade that happened March 13th that I believed stalls many of those issues as well as there was a parameter change I think came on the network March for 11th that lowered a bit amounts from five akt to 0.5. Akt so some of the Lively discussion that went on at the last providers meeting went into direct actions that have positively affect the network so much appreciated everyone's participation here. + +Tyler Wright: Is there anything that anybody wants to talk about related to the account provider that has been talked about? Quite yet on this call today. + +Tyler Wright: cool + +Tyler Wright: all right, if not then obviously everyone here is very busy. So I don't want to hold you. I know there will be a number of updates that will continue to happen and provider announcements in the providers Channel. I think aniluded to a GPU pricing page that will be added via the working group that cost website very soon. So there'll be announcements. I'm sure there again providers are continuing to come on to the network via provider incentives phase one and more relationships have been had since I know there was event set Nvidia GTC and a cubecon in Paris where conversations continued with various data centers all over that are looking to provide resource on the car. So again conversations are continuing there. as just a general + +00:20:00 + +Tyler Wright: All out and I think we've talked about this previously. But if anybody has any relationships with data centers in Latin America, Australia areas like that where you see kind of less folks generally on the map, even though again the map is continuing to get dots on it all over the world day by day. Please feel free to reach out to myself or somebody from the court team. We're actively always looking for kind of three tier four data centers in amongst many other regions, but certainly in those regions + +Tyler Wright: Cool, there's nothing else that anybody wants to talk about provider related then again. I'll certainly let you go. I would ask that everybody continue to push these messages that you see and provide our announcements out to again various providers that you have relationships with or connections with just to make sure that everybody is using the most up-to-date kind of upgrades and experiences. So again much appreciate everyone's time if anybody has any questions or comments, so they want to feel free to drop those in providers Discord Channel and we could talk about those at the next monthly meeting or between meetings. We'll have the Sig clients monthly meeting is following this in about 36 minutes, or we'll get some updates from + +Tyler Wright: the Maxes and some other folks that are looking to build tools on top of Akash, and then steering committee is planned for tomorrow. It's a normal end of months during committee. So just look after those couple of things. But much appreciate everyone's time today. Hope everyone has a great rest of their day and hope to talk to you all soon. Thank you very much. + +Andrey Arapov: Guitar lesson here on a good one. + +Scott Carruthers: Etc + +Jigar Patel: Thank you so much. + +Rodrigo Rochin: You guys. + +Meeting ended after 00:46:58 👋 +