 Okay. Hello, everyone. Welcome to the Blazar project update. My name is Pierre Rito. I'm the PTL of Blazar for the Stein cycle and There will also be other speakers. We have Masaito Moroi who was the PTL since the Ocata cycle and Tetsuo Nakamura and Bertrand Souville were both involved in the project So first Masaito will present the Blazar project Thanks Pierre. So I'm Masaito Moroi. I was running the PTL for Blazar project from Ocata to Rocky So first of all, I want to introduce what is Blazar project. So Blazar project is a reservation as a service in the OpenStack community Blazar project is created in 20,013 years under the name climate But it was inactive between the kilo or riveted to the pike cycle But we revived this project at Barcelona summit in 2016 and We became the official project since last year September from it's actually the Queen's Recycle and the Blazar project now supports reservation of computer resources currently two types whole host reservation or instance reservation in your crowd and This guaranteed resource availability for our future event as you want and The specific or the well known use case for the Blazar project is There is two use case first one is telecom NFV like OP NFV promise project and Another is large-scale experimental test bet like a in Camerion Camerion project And now we have a 10 10 active contributors in the rocky cycles And so we have a bit We are active projects in OpenStack in OpenStack communities And then this is how does Blazar works in your terminals So first one is a host reservation as I introduce the first one So really the host reservation reserve a number of hosts for the time period of risk like with with Blazar command if you If you type with a Blazar command physical reservation like a mean number of your Number of hosts you want to reserve like a mean to marks for and then Blazar returns reservation IDs Then blazar ensures that the project gets exclusive Access to these hosts related to the reservation ID and then when you launch your Instances in the time period the only you are only your instances Attached to the reservation ID is scheduled on the reserved hosts like this command and Next one another one is instance reservation in instance reservation you can reserve a number of instances with specific flavor size like this with Blazar command reservation and the vcpu numbers and number of Memory megabyte this gigabyte and how many you want to reserve in this size of flavor instances and then the Blazar returns flavor ID to users and In during the reservation time period you can use this flavor to launch instances and Then the Blazar ensure your flavor is Is not your flavor, but your instances with this flavor is rounded rounded and anywhere in Opens you are open-circuit without no valid host error And now this is the instance reservation only affinity Affinity and to affinity is supported for now because of some consulate in the current implementations, but we have now to get rid of this this list Consulate to in the stand shine or T cycle And this is a project of this session So I want to introduce you that the main features in now open-circuit rocky cycles There are five main features first one is a is for user features This is for the exit support for computer resource reservation Before the rocky release we you can't specify the exit name that you won't Aware you want to reserve compute host over the instances will be Reserved but from rocky cycle you can reserve you can specify is it name or is it? yeah, is it name and For we have two operators New features in the rocky cycle first one is a time margin for cleaning up free sources between reservations It's usually expect use for ironic support or some big big instances like blazer ensure that each user's to create their own Instances stick to the reservation blazer Remove or delete instances, which are when finished users reservations but in some of cases the instances No, sorry the Nova has a time to delete Instances so operator can specify 10 minutes or a few the number of minutes to have a margin between the each reservations and Second one is a periodic healing for resource monitoring the blazer project has ability to react the failure of compute host to reschedule the reservation on the Risk risk schedule of the reservation related to these failed host but before rocky cycle blazer blazer calculate whole Reservation related to the Host like if the host Goes down blazer recalculate from the reservation from now to the nine nine thousand nine nine hundred 99 years, but now blazer can blazer can handle these these problems like blazer only calculates one one days ahead what one week ahead something like that and Another the other new worthy development is to blazer fixed the some API response codes in rocky cycle because some of API response code was not Reasonable for every reason of the failure, but we We fixed these problems and of course we Yeah, and second wise they've stopped support in the blazer dashboard. So when you test The blazer project in your local you can easily test this feature from CLI as well as Dashboard with meaning horizon So next is our plan for the shine cycle So let's switch to the pf. Thank you. My say it up So for the Stein release, we are working on both user and operator features So for users, we are prototyping network reservation. So we are planning to offer users the capability to reserve things resources that are in limited supply such as floating IPs or Villain segmentation IDs or vx land for that matter We are also working on open stack client support at the moment. You have to use the Python blazer clients. We will you cannot use the open stack Unified Client and We'll also have a new API That's we call reservation candidate API and that will allow users to discover when resources are available for their reservations at the moment they have to try various parameters to find when reservations are available and for the operators we have a new API which we call the resource allocation API and This will allow them to monitor. What is the status of the reservations on their cloud? Which nodes are being reserved and which are free and we're also working with integration with the placement API and This will allow us to remove some of the limitations that we have currently in blazer and Make the scheduling more efficient Beyonstein we have lots of features planned as well. So as we move towards more production deployments We want to have support for high availability Integration with Nova Kotas is quite important as well at the moment blazer doesn't Integrate with them and so a user might be able to reserve more resources and they actually can use from the Nova Kotas point of view Compatibility compatibility with Aaronic is quite critical for some of the deployments that we have So we are planning to solve this as well Preemptible instances is something that's been actively developed in open stack for a while and it's a very good use case for blazer because The only reserve resources are kept idle and we would like to run preemptible instances on them and finally the Network resource reservation that we prototyping in the stein cycle We would like to make this production ready beyond stein in the tree to release and also work with the Cinder project to allow reservations of volumes and with other open stack projects as well so now Tetsuo will talk about our work with the placement API Hi, I'm Tetsuo. I am working on both blazer and placement project and Blazer is in stein's release blazer is going to talk to placement project directory and blazer resistors reservation resources to placement and Nova will refer the reservation reservation resource class when Nova wants to deploy virtual machines and by Using placement we want to enable several things that this Blazer will no longer depend on specific number filters such as these three filters and the important thing here is we We will support a reserving multiple instances on the same set same host Intentionally that means we will support Affinity policy is true and we are targeting these features in stein's release and We want to use placement for host reservation to Be on the tea release cycle and we will and then we will no longer depend on blazer filter Okay, that's that's all from Cross project work topic. Let me switch to patron Thanks, Tetsuo. So if your user Interested in trying or even intending to use blazer There are plenty ways to do it and starting from this week So we recommend you to attend the forum sessions. So tomorrow we will discuss about extending blazer reservations So to new resource types like networking and storage So this will happen tomorrow and Pierre the current Blazer ptl will be the moderator on First day as you may know There are several projects right now that are planning to consume the placement API like cyborg like neutron besides blazer and Nova and Tetsuo will moderate this session and we will discuss whether it makes sense to have a common Python library to consume placement API Of course, we also welcome you to attend our weekly meetings that is held on Tuesday At 9 a.m. UTC on OpenStack meeting art RC channel Oh, of course, you can also always talk to us directly on three notes And what is also quite important, especially if you are a user Please do not hesitate if you find any issues to report bergs directly on the blazer Also quite important, of course, we are welcoming also new upstream developers So first thing that you would need to do is of course to read our contributors guide. You have the link here But also if you have some Python coding skills, it's always good to have more eyes on The actual patches. So if you can help with cold review or even help to fix actually some bugs so Helping with bug triage or even improving the dogs and Improving the test coverage, of course, you're much welcome And as I said, please talk to us either this week or later online on RC or doing our Online meetings on Tuesday So once again, thanks all for coming and we have a few minutes. I guess for Q&A Yeah, so when I was saying Nova Kotas, I meant Kotas applying to instances, you know, however they implemented So it will make sense, of course to use the more modern approach, which is the unified Kotas U.S. markets and we have very often a problem to integrate a new project from OpenStack because there's no comparison to market typically other services. It's a plan that you also integrate in your Support and helping things that you make market comparisons or learning comparisons. So the typical customers will be having a more than one cloud ecosystem Public or private, VMware or AWS, they have a completely different wording to OpenStack So simple scale scenario from Functions to Service. Everybody name is Functions and we in the OpenStack Foundation We name it MeteorSore. It's a complete visit plan here in this project lesson to make a market Merlin learning that also other companies which are actually using Services from other clouds that we can invite them to participate So reservation is quite unique actually in lots of public clouds. You don't reserve resources in advance because they make sure that they have enough to run their own demand. Are you talking about what they call reserve instances? Because I'm aware that in AWS, for example, you can buy an instance for a year or several years but the service provided by Blazare is a bit different because you reserve only for specific period of time that you actually specify. That's a good point. We have to map to the existing platforms which are there. It's fine that we start every project new but when we take every time new wordings or things for context there is a learning place on the market. Then it is really hard to explain a customer. Yeah. A simple example, Volkswagen Group. We can switch they mostly away from AWS when we have it's one of my main customers. And I know that very often they decide it's versus OpenStack because they don't understand with some reservation service that all the reservation service is existing on OpenStack. Because the assertion is learning what they are knowing. It's only a question when you say no and I will place the target on the demand for it. Because this is a problem that we have in every project. If you are looking for something in OpenStack or this could also be done in some other projects. The PTIFD Promise Project which is very much related to this work. So such a mapping before the first approach right there. I understand your concern is a lot about terminology. There is different communities. Yeah. It was very much a cause I know in some projects for example NOVA. They make this mapping. And it helps extremely when first of my architecture and my evangelism team which is located worldwide in more than 40 regions when we talk to customers. Because everybody from a technical side has a preferred platform. It depends on where it's coming from. And when we want to have for example a server more successful then we have to make this mapping. It's an initial work. It's nothing what you have to do 10 hours per week. It's only one time. Thanks for the suggestion. I mean we look into it. What I can say is I think it's quite unique among the cloud ecosystem. So there might not be anything in the main public clouds that look like it. I think we have to stop here unless there is another question. Thanks everyone.