How Passage Reduced Costs with Akash Network: Virtual Experiences & Meetings
15 Jul 2024 · 9:52
Sebastien Couture talks with Lex Avellino of Passage about the company's partnership with Akash Network, a decentralized compute marketplace. They discuss how running virtual experiences and online meetings on Akash's infrastructure cut computational costs, along with the practical challenges and future plans for the collaboration.
Speakers
Topics
Transcript
you guys recently announced a um I guess a partnership with aosh and from my understanding um a lot of the 3D rendering in passage is being done on the aosh network can can you explain yeah you know how with a caution how it's um better than you're using centralized servers or or other sort of computational resources yep absolutely so um one of the thing one of the decisions we made from the outset right we've it's it's really wild how little we have compromised on immersion or accessibility I mean in some ways this is just as accessible or more accessible than like YouTube Google meet it's one
click and you're in right I mean it is dead simple um and then immersion we we have the best possible Graphics now you go and look at something like uh the games strange Clan wah wa like they can be more stylized more cartoony but then you look at some of the environments like def dojo and it can be extremely realistic all within the same game engine so no compromises there well you have to compromise somewhere right and the compromise for us has been one our sanity in building out this wild system but then two uh cost right it is more uh computationally expensive to host these experiences than anything else right I mean I remember when I used to think when we were first doing passage I used to think that the video hosting was
expensive like doing the whole webcams um but then we're still rendering everything in the browser so your computer was bearing the comp computational cost not us uh that was early on in in passes when we launching you know our event with Amazon for instance um but um that doesn't even uh the computational resources of of video feeds um doesn't even you know shake a finger at uh uh the computation of 3D worlds right I mean we're talking each user on passage occupies about half of a GP now these can be anywhere from like commercial for for those who care commercial gpus to high-end gpus so like we have a good range of cost um
depending on we'll basically have right now it's it's all one tier of GPU that you're getting when you jump into passage but eventually we'll probably have like tier one and Tier Two Worlds some worlds that require a little more GPU power to run um so that people can get the most out of the the hardware um but so so it's not like the most expensive gpus that we're using for passage but you're still occupying half a GPU which for anybody who hosts a website or whatever like you know that like maybe you use a couple CPU threads for your website for the whole month for all your visitors right um the computational demand of passages is Big right uh imagine you got to use half of this big glowy monster for every single
person you need one of these for every single for every two people who are going to be in an event so that said it can be expensive uh to run now we our belief was basically that um and this was this was you know four four years ago that we started building this um so our belief was that it's okay to compromise on price that can be the compromise because this is the one thing that will change fastest right and I think we've been affirmed in that people's ux journey is not going to be expedited it's going to be slow and people are not going to adopt a new user experience quickly but the price of Hardware will adapt quickly and it has I mean I believe we've seen as much as a 60% drop in GPU costs um
from the time that we were building this in 2020 um so we've worked with both gcp Google Cloud as well as uh AWS right we've used a bunch of different uh uh Cloud Solutions uh passage is cloud agnostic we can leverage any architecture um but point being we we've seen the best and the worst pricing in Cloud compute and it's gotten way better since 2020 because again there's been a lot of push for um GPU infrastructure for you know Ai and for all sorts of different applications that said it's it's still relatively more expensive than any other kind of Hosting uh so this is where Akash comes in Akash is the first really open market for gpus right I mean in
terms of people offering gpus for uh real world product use you've got as your Google Cloud AWS that's it three people right there's like three people competing with each other and so it's a little bit of a rigged market right you can't really call this an open market uh but what akos does is it allows anybody all over the world to contribute GPU compute to their um decentralized Cloud Marketplace and anybody can buy it and the relationship between provider and the um the the person leasing the hardware um is governed by smart contracts right so the beauty of that is there's censorship resistance um there's a bit of security just in terms of
infrastructure it's actually more reliable in a lot of ways than a AWS has been down at times when aosh wasn't um so the up time's better but then um there's a lot of advantages to AOS but the biggest advantage of having an open market place is true price Discovery right the the cost of a GPU the same GPU on aosh versus AWS is I mean up to 10 times cheaper and for us personally we've seen cost reductions integrating a cost of of 70 up to 70% for the gpus that we're using and there are gpus you can get on a c you just cannot get on AWS because there's so little availability and you've got to go through a whole Pro a permissioned process to get approved to get these
gpus all that nonsense with aosh in one day in one hour you can have your developer you know get your your akt and set up your your lease and you're you have access instant access to Hardware that isn't even avail available anywhere else but for a price that's 10 times cheaper so that's wild that's super wild um and we've been using that so we've been building putting more and more of our systems onto a c um and how's that been like I mean what's the what's the integration experience I know you're not you're not you know on the technical side but how easy is it to integrate with a c and leverage those resources for what you guys are doing
yeah I mean I would say easy right I mean okay it's becoming easier easier depending on your use case because there's more people building tools for leveraging a Kos compute uh so you know you can go and launch a a WordPress website and aosh in just a few clicks using some of their uh Native tools built by teams like cloudm right um so it can be stupid easy um now I am biased because we have some really smart developers um our CTO is is brilliant um and so I'm I I am definitely we're definitely skewed on the side of being highle developers so you know for just your kind of average run-ofthe-mill full stack Dev there might be a learning
curve um you've got to be able to work with um kubernetes there's some contain containerization stuff you have to understand but the beauty about a cost is it's built for scale right now that does make it in some sense a little bit more complicated than it could be but for people who have big Enterprise applications it's perfect um there's so much really really good architecture um built into the aash network so yeah I would say uh any barrier that's there for your kind of run-ofthe-mill full stack Dev um is a good one right it's one that's going to uh create sustainability and scalability so yeah i' would say overall Dev
experience is phenomenal you know the community is phenomenal there huge open source community building around Kos everything on a cost is open source so really really easy to contribute to velopment a lot of people building tools for aosh so I mean jump in the Discord and you'll get your questions answered stat I mean really really great community building around that so that really helps uh develop developer adoption cool no I mean it's it's a great use case I think also for aosh to you know AOS I think has made this sort of pivot into the GPU uh Marketplace um you vertical we had Greg on several months ago where we we talked about this uh you can go back and listen to that episode if you're interested yes
he's brilliant uh yeah no absolutely and um uh and and this is just another use case I think that demonstrates uh just how powerful aash is as a Marketplace for compute oh yeah not only for use cases outside of crypto but use cases in crypto as well it's a yeah we did a booth takeover of the aash booth at eth Denver and um it was so easy to sell people on this right so we're talking to people about aosh and they're like oh you're from like no I'm from passage and we're saving 70% on our GPU Bill Boom sold right like it's just wild in terms of um anybody who understands the need for GPU resources can immediately see the value proposition um so it's an easy
sell and um yeah we love having those conversations because it's been huge for us and we're just trying to move more and more of our services uh or you know infrastructure stack over to aash so right now we have a little aash button in the dashboard where you can launch event on aosh versus AWS and it's it's my favorite feature so far because it's kind of like the instantly save money button [Music] [Music]