I have. I don’t claim to fully understand everything but two things caught my eyes.
It is a congestion tool. But I fail to see any congestion. Like my CEO said, congestion tools are for people with not enough bandwidth
Does Youtube see a 4% improvement. If they only can improve by 4%, why should I even bother with a STORJ node?
So no, unless you can provide me any reason why that should make an impact, I am not even considering it.
As far as I am concerned, you could also tell me to whisper nice words to my HDDs, because they will then perform better. You know, like some people claim you can do with plants.
I read your answer, however, to my eye it looks like a poetry. I have no incentive in convincing you to anything. To be honest, I think that you are misunderstanding those papers. I cant commit more time. Sorry, really have to go.
This is a curious thing here, because BBR actually applies here. You need to spend more than 5 seconds to understand the details though, sorry, so feel free to skip the explanation below.
Everyone in the Internet uses some sort of congestion algorithm, it’s not an on/off switch. Instead, it’s a matter of choosing which congestion algorithm is your node using. And why this matters if there’s plenty of bandwidth? The fact that a conservative congestion algorithm can prevent you from using that bandwidth!
The key thing is that what is called a congestion algorithm is basically a tool to define a trade-off between usable bandwidth and the risk of congestions. Why is that? Because the primary way of controlling congestion is by limiting bandwidth. The more pessimistic congestion algorithm you use, the slower your bandwidth performance will be (usually at the beginning of the connection, so called slow start). BBR is more optimistic in this sense than the usual defaults—it assumes congestion is less likely, finds a better way of detecting uncongested links than previously popular algorithms, and as such is capable of taking advantage of more free bandwidth.
This is why BBR is commonly advised—not because it controls for congestion, because all networking stacks do that. It’s because its way of controlling congestion is better suited to modern networks.
No, but this is a hypothesis that can be tested. Have you tested it, or maybe already dismissed even a possibility that BBR will affect Storj nodes based on a single googled sentence?
Yeah. And I can also test if yelling at my drives or how the moon phase affects STORJ.
A lot of things can be tested
No. I tried to understand it and did some research. A few things came up during that research.
For once, the gains are very minimal. And for latency the gains are even smaller.
Second, BBR is not without its critics. There were some reported instances, where BBR even worked worse (heise.de). Hence the version 3 in not even a decade.
So to summarize, you guys still failed to even conceptually explain why BBR could have an impact on STORJ.
Ignoring that BBR probably already is active because the host Debian 12, since BBR has absolutely 0 to do with ZFS or filewalker, I could also test how fast my VW accelerates from 0-120km/h.
but this is a hypothesis that can be tested.
Before we do that, maybe we should test first if congestion or packet loss is even problem to begin with. Because if it is not, testing would be pointless.
I would go even further and say it does not help any STORJ node with more than 50mbit bandwidth
But why do we reverse the burden of the proof?
Aren’t you supposed to explain why it could help, at least in theory?
But I like learning new things, so if you have the time (@s-t-o-r-j-user only has time to say that he/she does not have time), I would love to look into it.
To me it seems like IPv6. I love IPv6 and hate IPv4. But I still know that making my node IPv6 does not help. Why? Because basically customer uses IPv6. So the reduces latency IPv6 offers is nice in theory, in practice does not matter.
So my first question would be, even if BBR works like a charm, best case scenario and I really would get that sweet 0.2% reduced latency, is it even supported by customers?
Let us assume, it is supported. How does BBR actually apply to them? As far as I understand, BBR is great if you try to get something from a single source, and there is some congestion. The congestion is only detected when the bottleneck is already overloaded, leading to large delays hurting interactive applications. BBR can help reduce latency by not doing that.
But how does that apply to STORJ?
One theory I could come up, is that because BBR in theory should have a slightly faster start, that you would win races against other nodes. Basically you get a very small head start. That sounds nice in theory, but then again, there are probably ten other factors with a higher impact.
Hope I answered your wired questions, nevertheless, the main reason is the form of communication, I dont know why is that, why I got this kind of a feeling that its indeed a bit wired.
And I am sustaining this statement. And again, I provided you with the explanation in the form of two papers. And inside those papers, there are links to additional papers. Don’t take it personally, but I’m no longer interested in continuing this discussion. And please do not refer to me anymore, nor using the first nor the form of the third person. You won’t hear any reply from me anymore. :- )