Hello
I currently have 2 internet links (whit different public IP), I would like to know if I can take advantage of this situation on my node
1 link I have almost exclusively for the node and the other I have for personal use and some websites that I own,
Would adding this second link increase my reputation?
Is there any way to limit the maximum amount of bw that can be used in one of the links?
Maybe i need to create a second node to use the second link? (and if so, do I have to request a new authorization token?)
1 Like
You can and in that case you need a new token.
I have two links, but they are different capacity (the second one only has 12mbps up), so I just use the main one for everything and if it fails then everything switches to the backup link.
So long as your single node isn’t saturating a single link, there’s not much benefit in attempting to aggregate (merge) the links. @Pentium100 strategy is good for maximum uptime.
The most benefit in the current network state will come from running a second node on the 2nd link at the same time.
You’ll need to use 3rd party traffic shaping programs to limit the traffic.
the problem is in the second computer i don’t have the minimum ram, i think add more swap space but i don’t know how much affect the performance of the node
I wouldn’t rely on swap space. It would be pretty bad, especially for the DB. You can run multiple nodes on the same host, but on different connections. You’d need a seperate NIC, and management of that might get more complicated for maintenance. But, it’s another way - if you can’t afford RAM.
I agree and I would be doing that if my second link had better upload.
Actually, the best option would be to run two nodes on the two links, but configure them in such a way that if one link goes down, both nodes use the remaining one.
As said earlier, I suggest you to run one node by connection. With two separated connection, you actually decentralize the nodes with your own residential connections, which is the way to go. The aggregation is a costly method if it works well, because the cheapest and easier methods I found are only for personal use for websites, without peer to peer or ability to open ports, so it was useless, domestic routers are not made for this and devices for aggregation are awful to configure and costly (and it could not worth it all for running a node). For the real usage at this time, even with testing data, Storj doesn’t need a huge upload bandwidth most of the time (you’ll see by browsing in the forum that most of the gigabits internet crunchers get only 30mb maximum of upload most of the time, and most of the time, here, there are no upload at all), so if your upload is better than 15 mbps, you might be totally fine to run a node on this.
Here, I have two separated VDSL 2+ connections (100/35 and 85/20 on second one because this ISP is briding contractually the upload for residential users (I hate him haha)) from two ISP (like you I guess) and two routers, but on one system by using Docker Desktop 2.1.0.5 (not the 2.2.x.x, you could run into troubles), one in each container and I changed the port and the local IP in my second node to fit to my second local network and avoid port conflict. And it runs quite fine here on Windows 10, you can observe the traffic activity in the task manager on each ethernet interface, which matches to the realtime logs (when you see a bunch of uploads/downloads, don’t forget to switch in your head by downloads/uploads because the logs show the customer view when he downloads/uploads data). My 20mbps upload is visibly totally fine for downloads pieces requests, smooth most of the time, I’ve like 90% of success rate, which is totally fine.
If you don’t have two ethernet ports, if you’ve a PCI Express, you’ve maybe the possibility to add another cheap ethernet card and connect it behind the best machine you’ve if you would prefer attach drives by SATA and check locally the two nodes on a second screen with two Windows Powershell windows and a view of the drives/network directly from the built-in taskmgr like me. Its my main machine, so I’m often there to check it and I don’t need to use my local network to do so.
But, in any case where you’ve issues with this setting, you can still use another machine or buy a mini pc linux machine like the Odroïd computers that some users use there, like the Odroïd HC-2 having a SATA connection and dedicated enclosure for a drive, which is kinda cool!
Just get a Mikrotik. Or build a Linux router. It is not that difficult to have connection failover. Well, maybe not on the cheapest router, but you should not be using the cheapest router anyway.
In my case, the 12mbps upload of the second connection is too slow IMO. While in theory I would get more data, the second node would have problems uploading it and upload the main source of income, not cold storage. So, I just use that connection as a backup (I am pretty sure that if that ISP ever looks at my traffic graph they will be confused).
If I ever get another connection with better upload, I will be setting up a second node, with the idea that if one connection fails, then the nodes will use whatever connections still work. IIRC IP filter works only on current IPs and once my connections get back up, the nodes will not longer be aggregated.
1 Like
What second Computer?
Oh yes, I understand Pentium. With 12 mbps, it looks a bit too tricky for the upload traffic, I admit. And yes, I never tried to do that kind of router, but I watched some videos and docs when I was thinking about it. I still think it’s not so easy for the common user and if its especially for Storj, and I heard a lot of difficulty and stability problems about this, so it ruined a bit my wishes to try this; but that could be a great solution to combine two weaks connections, especially two ADSL connections if the sum of both could match to minimal requirement of Storj. Yes, Failover could be of course another security solution, especially for a second connection not really usable for a second node.
But when you talk about failover, there is a problem. I’m talking about two separated traffic nodes, they never share the same public address, so they receive the traffic individually from the satellite, because that’s exactly what it is, two individual connections, never shared in between them (except for myself). From my perception, you can’t use failover on the down connection for the node using it, because you can’t share traffic with the other node, or the satellite will assume you’ve only one shared network, or you could be penalised when satelittes detect two nodes using the same public ip during the failover. You could lose and reduce by twice your profit. Am I wrong ? I don’t know how the sat. reacts when the failover is finished. Because if the filter is applied once for the lifetime of the nodes when a single public IP for the two nodes has been triggered by the system, that would be a big downside for two connections with failover in between. But if it the filter applies only during the failover, so its perfectly nice of course, and I could consider it myself.
But if its this not case, it will be more profitable to keep the two connections as it, separated without failover in between, with the same range of ISP address behind each dns address and accept the downtime of the node (or only change the IP of the down node if the downtime becomes critical to avoid disqualification). However, for two unstable connections, I can also understand that the failover could be safer if the downtime due to ISP reaches 5 hours and disqualifiates the node. So on the end, it’s about the user, connections state, infrastructure and choice for sure. But, the nodes have been normally made for home users basically, with their residential connection, people like us having two connections at home is something quite exceptional (luckily here, we had already a free 80’s copper line reserved for relayed phone in the prehistory, so technical installation was “free”, so I use it as my main connection, the other is still used for TV, not available with this ISP, and as failover for me very rarely, because the monthy uptime is generally close to 99,7% in worst cases)., most of them have to deal with one connection. And Storj could still think about how they manage the disqualification, I read its a huge debate, but here, 5 hours of downtime only from a single ISP never happened until I remember, but I worry it could happen elsewhere in the world.
The “failover” is working but only for myself on Windows 10, which is automatic. When one connection fails, I’m connected on the other one. But, to maintain and be connected again on the main connection everytime I restart or be disconnected temporarly and switch back to this one, I have to use a lower value in “Metric” in the properties of this controller or in the shell. In that way, it works fine.
pretty sure you can setup two nic’s to load balance, its fairly common server practice, because of limitations or prices of cards or infrastructure, and then ofc the added redundancy, it just makes sense to use such a feature.
with something like that setup correctly using dns names instead of ip’s you can route multiple paths to one application.
i would say it’s easy, and in theory it should be… but theory and application is not always the same xD
the real question is if the tardigrade or storj will accept using an unresolved dns name instead of an ip, but in case it wouldn’t you would need to have an singular ip offsite that will receive the connection and then load balance it between your two online ip’s or use dns.
… and really why would you need it, storj won’t pull and the bandwidth anyways… or my 400mbit still haven’t broken 110 peak and the long term avg is about 30-40mbit.
if you just want the redundancy you can simply setup failover… its much less work and trouble i bet.
or two nodes obviously… where each would failover to the remaining ISP connection, that would actually be a pretty good solution… you get the performance of two isp connections and you get the redundancy and its easy to setup… ofc you don’t get the massive possible peak output on singular data pieces, but really… considering how easy that would be to make work compare to the other stuff… then that would most likely be my choice.
1 Like