Just read this article, which might be of interest.
Their own post here.
Nexus mods seems like a very good fit for Storj, especially when dealing which such load spikes from globally distributed users. They would host mostly larger files. The article contains pretty specific figures about the peak loads they are dealing with, perhaps Storj can look it over and drop them a note if they think it’s a good fit?
Both links make it sound like they absolutely know how to handle more load… they just can’t afford it. They likely can’t afford an enormous amount of egress charges with Storj’s pricing either: but I guess it’s possible?
or can they? @bre @Alexey
As Storj is pay as You go and sort of instant CDN?
v.s
traditional cloud, where the scaling is optional and is the most costy part i guess?
It sounded to me like that don’t have scalable architecture and they can’t afford to scale up permanently for momentary peaks like this. Also traditional architecture deals with bottlenecks that can be avoided using Storj. But I’ll leave it up to Storj to consider reaching out and Nexus Mods to consider a switch. 
We have sent this to the team for consideration, thanks!
This is really good, but going forward the prediction is what most valuable, I mean, if there is some sort of events that would make some company need download stuff, seeing in advance would really help 
If Storj had built this: Better Performance for "Hot" Files · Issue #11 · storj/roadmap · GitHub
I think some of that may have already been resolved with the edge services now running on a distributed architecture as well. Though scaling the amount of erasure encoded pieces with demand might still be relevant. This has been mentioned as a future direction in the white paper from the start, but unfortunately hasn’t been implemented yet.
Yes it is unfortunate, because it could be a fantastic selling point for such cases like the one you have linked.
This makes me smile. I remember a project that just bought a handful of cheap ±$10/year VPS packages around the world with unmetered traffic… and used them as BitTorrent seeds. Pretty much linear scaling… and it was all they needed.
That Bram Cohen guy was onto something
You’re probably right. That’s why I posted with some hesitation and left it up to Storj.
But… If I’m honest, this should be an inspirational case to strive for at the least. Even if the implementation isn’t currently ready. 190Gbit spread over all nodes should come down to less than 10Mbit per node (taking into account some margin for long tail cancelation). Now that’s a naive calculation which ignores all the challenges you outlined, but that is definitely doable on a truly distributed network.
As for S3, sure that is the most likely way they would want to implement it at first. It’s definitely not ideal for this case. What I would like to see happen is that the Gateway MT is implemented in a truly scalable way by Storj. Gateway ST can already run completely independent from the satellite, so there is no reason the same can’t be done with Gateway MT and load balancing. Just spin up instances dynamically based on demand.
Of course the ideal solution for these scenarios is to have a native implementation in the browser. @jtolio did discuss that a while ago on the forums and mentioned that it would be complex. But it would broaden the types of use cases Storj could support by a lot.
Pricing can be changed, especially for high volume customers. And after the recent changes, node operators don’t get much money from egress anymore anyway. And if a native client for browsers would be available, Storj doesn’t have to pay for the required edge services anymore either. I would be perfectly fine getting paid less for native CDN like use if it means Storj could onboard such clients.
Either way, if not for a client right now, I think this is still a use case to strive for long term.
Remind me of this project https://webtorrent.io/ (it was a bit buggy when the number of peer is low last I try it).
Not really. It’s more like you’re trying to make 2 deliveries about 80km away in different directions within an hour with your two Vespas and two delivery people, but the customer requires them to be tied together. 
I’m just arguing to remove those ties.
It seems like you were reading and responding at the same time, since you seem to have changed your attitude at the end of your post.
I’m not arguing that it’s possibly now, I’m merely arguing that the number of nodes there are currently, should be able to support this use case, if other bottlenecks are dealt with. It’s frustrating to see that solid core be limited by the chains of the edge services. And as @kocoten1992 showed with the webtorrent example, there are ways to do something like this in browser, which would eliminate most of the bottlenecks. (Though I’m unsure if other bottlenecks on the satellite end might pop up.)
So yeah, I’m just advocating to unleash the chains that hold native implementations back. Let’s make these nodes blaze!
hi.
i don’t think any S3 is a problem, the Website can offer a file installer, that customer downloads first, you know, You want to get that 130GB Fallout mod, and You first download an installer.exe like 1MB, that enables Your PC to establish a storj native connections from Your desktop PC and get that 130GB mode from Storj nodes.
i have 0.3 Gbps and i don’t mind they saturate it, i don’t use it anyway, and i have to pay the line anyway.
Oh please no… this is not a solution and would be universally hated by their customers. People don’t like installing things or extra hurdles when they just want to download something. It would be a little different if it was an optional modding tool with broader purpose to actually install and manage your mods with this as an additional functionality, but to have it required for downloading mods from the site is a no go if you ask me.
¯_(ツ)_/¯
maybe they already has such tool, idk, never logged there,
was a popular solution in many drivers software websites.
Yeah, the difference is drivers already require an installer to begin with. So to have that installer download the files too is not an additional step. Mods tend to be just files you download and place in a folder.
That said, I think offering Storj as an integration in installers is actually another excellent use case for Storj. It’ll make software distribution a lot easier and it would be much easier to avoid the S3 bottleneck that way. So good catch! I’m just not sure if it’s a good universal solution for this case.
You are wrong. Each Gateway-MT instance is spreading this over nodes. We just need more Gateway-MT instances to handle the load, as suggested by @BrightSilence
It was, since it was exist. They did not survive - that’s another outcome. However, implementing uplink in a browser is a challenge. Could you please participate to solve this challenge?
I do not know, but it will be rewarded. You may try to participate.
If you would offer a PoC, I can assume that you will not leave us without a reward!
I would offer to make this challenge as a regular reward program. We have some feature requests in our GitHub repository too.
I asked the team for the reward details.