How to Deploy on Fluence Cloudless Computer I Alexey Proshutinskiy and Dmitry Kurinskiy

I'm Alex. I'm engineer at Fluence and today I'll show how it's easy to deploy on Fluent Cloud Platform. But what is Fluence? Fluence is decentralized cloud and we do it because of a key problems with centralized and traditional clouds. It's a vendor looking because uh if you build your like whole application project based on one uh cloud uh it's really hard to migrate from one to another and it's like really really huge uh cost in hours and work.

Also a lot of clouds traditional clouds charge a lot of margin on top of the actual computing services because they provide a lot of services and stuff but it's really high cost and like it's most uh biggest expense of some companies just hosting. Also, you have censorship risks. If you build all of your product based on one cloud and you get banned because of something with le free or crypto, you just ruined your business and also sometimes you just don't know where your data hosted. Uh how much do you pay for them different stuff? So because of uh these problems and our hours we built fluence and what is fluence?

Fluence is a centralized computing protocol. It's a platform and uh it provides uh CPU cloud and uh GPUs, containers, uh bare metal and VMs and uh why it's cheaper. We have uh agreements with data centers tier three, tier four which like provide enterprise grade hardware and also we don't charge much on top of the actual cost of hardware and also providers incentivized to join the network and uh they receive all rewards for idle workload. So when they wait for like actual deployment they receive something influence token. So it also like make price lower also because of uh like uh our platform and uh our like uh general API you can deploy to multiple data centers in different location and if some data centers like don't want to host your data your workload you can move to another one without changing anything in your code you just use the same uh API we used before and uh about our case features.

So we are really cheaper when most of the clouds like AWS and Google uh you have no vendor looking because it's like network with different data centers and providers also we have like the same data centers and hardware it's certified it's enterprise grade we have multiple locations in Europe and USA uh we don't charge any fees for uh regress traffic and uh Uh we for now we provide like full stack of infrastructure. Uh we like recently released our aggregator for containers and bare metal and GPUvms. So about hackathon and bounties we have two tracks. First one is you uh use our GPU containers API and uh basically your product uh need to be created on top of the like uh necessity to deploy some workloads for GPUs for like users of some stuff and utilize like API for it. uh basically for now API available not only for containers but also for like bare metals but bare metals is too expensive for hackathon and it's like better to use VMs or containers also if you like run your workload uh with uh GPUs from fluence you also can host your back end on CPU cloud and also we provide like free credits for hack hackers to make it like free and uh We have three teams uh in this track and uh also you can like uh apply to another tracks of our partners because we're like too general.

You can build anything you want if you like meet some compute. So apply to our partners bounties and yeah and the second one is CPU cloud. So we have CPU cloud API. Uh we have two teams on it. If you need some compute that doesn't require much of AI stuff and like web scrappers, you can do it with CPU only.

And uh about demo. So we have a console uh just to show you how it looks. So we have two tabs basically CPU cloud and GPU cloud. uh you can just uh use it as a like to play with uh uh platform overall. So we have like multiple locations you can use uh any hardware from like 2vcpus up to 128 CPUs.

You can like uh specify how much storage you need and it's kind of like really easy to deploy. Uh so let's just try to deploy something. So you can uh also deploy any custom images on top of it. You can specify ports. So let's like launch one in Germany also sh case.

So either you use uh API or use uh console in uh instances you will see uh in some time your uh instance. Yeah. So you can check what's deployed and where to find them. So you pretty have like all information provided here about your uh VM. Uh let's look at the previous one.

So uh I deployed it before the workshop. So you can see to connect to it you need to do SSH and uh you will have like instance working for you. So but but for hackathon you need to utilize API. So for API for CPU cloud you have like proper documentation. You can look at it at.

dev def also where is a link in bounties so you can check how to make anything to deploy to manage your like workload and stuff and to use API first you need to create API K in settings so first you create API K and then you use it as a bear token to make calls for your a for your API yeah this beer token. But for GPU cloud, for GPU cloud, we have uh really freshly deployed API and it has uh only swagger. So it's not so easy to understand how to deploy but I will create really short doc about it. So it will be easy also come to booth to ask. So basically for containers you have most two two most important uh endpoints.

One is two plans. It's which GPUs you have available in network and it looks like one second. Yeah. So it's plans and it looks and it looks like it. So you have like a lot of plans with uh pricing with uh hardware requirements and stuff.

So if you automate it and include in your project, it will be like good for a bounty. So let's deploy some some of it. Let's deploy this one. And for deploy you use uh this comment.

So I have something prepared.

One second.

Yeah.

So we just change one ID. And uh we have an instance it should be appeared here. No not here. So like previous instance which we launched in CPU cloud is ready. We can like also check it and uh yeah it's up and running.

have two vCPUs and uh let's check GPU one. So it takes time to deploy it. Also sometimes it happens when someone ran before you and uh some instance can fail. You can track it via uh this this one or these APIs. So basically we should wait a bit but maybe it will take some time.

So we have not much time left and uh if you want to try console or if you want to uh do some bounties you can uh scan this code or come to our booth. We also have this code. You put your email, you log into a console and like in couple of hours you will receive free credits to like try out our platform. Also come to ask any question about like your project ideas and stuff. Also we available in discord whereas like if global wenosirus channel you also can ask.

So if you have any question while we wait oh it's active so we can we can try it. Basically, it's Debian. So, uh, so we can try it. And it's 2349. Okay.

Oh, not it's root. So yeah, now we have an instance ready and uh one second and we can check uh about our hardware by this command. Okay, so we have our GPU ready. You can use it. You can like deploy any container you like for your bounty.

And that's all basically let's tweet it. So anyone has any question about bounties about platform and stuff we have like five minutes.

Yeah. Okay.

So like if you build some project uh to uh apply to a bounty you need to use like API to deploy some stuff on fluence. So you can host uh your back end on our cloud just like via interface but uh to be illegable you need to have a product which requires some on demand compute. So for example if user buys some training data on demand you can deploy from your back end on fluence to like spin up instance of GPU run some workload and like uh shut it down. So it's good. So you use API, you like create instances, you shut down instances, it's good.

I don't quite get your question. Uh so basically uh your project should require some of the compute on demand. So you have some back end running. It's okay. It waits for like users to come and pay for something or like rent something anything and if user do some like action like uh buy something you uh from your back end you using fluence API deploy for example GPU instance for this user with some custom image or custom container or something and it do some workload it shut down.

Yeah, that's a project but like uh you can get ideas from different partners. Uh basically you just need to utilize API. So like just deployment of your back end is not enough. So yeah. Yeah.

Like credits is enough for CPU cloud for whole hackathon also for GPUs like simple one too. Oh I am already finished basically. It's okay. Yeah. Yeah.

So any more questions? If not, thank you everyone. Come to our booth. Also, we have Pre