What are the plans regarding the identification of storage nodes vs. public IPv4 addresses used for node? In my infrastructure, I have the capacity to run several storage nodes, totaling some 100 TB of space, over several physical servers. However, IPv4 addresses are a scarce resource, and the ISP may not be able to allocate many of them. So far I have been running a single node on a Linux VM, but I am tempted to give a try to the Windows installer and run another one on Windows. Currently, I do have just a single public IPv4 address which I can use to route traffic to my storage nodes.
I think this could be quite a common scenario. I’ve read that Storj interprets storage nodes running under a single IP address as a logically single node. I believe this should not be the case for production in the long run. Storj could use the IP address as an indication of a physical co-location of such nodes, but the life-cycles, availabilities, reliability, and performance of each such a node may legitimately differ.
It’s kind of already the way you describe it. Saying the satellites treat nodes on the same IP as a single node is an incomplete shorthand for what’s actually happening. Each node behind a single IP still has its own reputation, so availability, reliability and performance measures are still node specific. However, when the satellite is picking nodes to receive data, it will pick assign a single IP with multiple nodes as many pieces as a single IP with one node. In short, during node selection they are treated as one node, but for all other purposes they are still separate nodes on the network with their own identity and reputation.
This is all done to ensure decentralization of data and making sure there aren’t any single point of failures popping up, so I doubt that’s going to change any time soon. Unless you have different physical locations with IPs in different /24 ranges, it’s not very likely you’re going to fill up that 100TB.
5 Likes
Great, that sounds very logical! From what I read, I understood the role of IP addresses in node evaluation much stronger. The way you describe it I think is the correct approach.
1 Like