clearly, egress is most important for SNO payments.
Nodes with 1 Gbit/s uplink seem to have a clear advantage compared to nodes with 50-100 Mbit/s uplink. Is it true that fast uplink nodes will statistically speaking more often win the upload race and therefore have a better egress / ingress ratio?
If yes, home operators over DSL will have a disadvantage running storage nodes compared to professional users operating a data center
if your node is 100GB then maybe your 1 Gbit connection is prooven, if not big nodes, then iven 20-30 mbit is OK as they not have lot of data and big bandwidth to upload.
Winning an upload/download race mainly depends on your internet latency (also determined by the distance to the uploader/downloader) and how fast your node can process the incoming request. a 50MBit connection would still be sufficient.
Also I think that datacenter nodes might have trouble because they share the same physical location with hundreds of other server that could potentially run a storagenode. Since the goal of STORJ is to be geographically distributed, those nodes in the same datacenter might be seen as a single node on the network and therefore share their bandwith, giving them no advantage over a node at home. But I’m not sure the ip filtering is already able to filter datacenters.
@kevink Although you sure have a valid point that a lot of nodes could be in the same datacenter, servers in a datacenter might still have an advantage under certain conditions.
If I think about the routing of the ISPs in Austria, almost all ISPs route their traffic over Vienna independent of the users location. That means if you would be a customer and your neighbor would have a node, the traffic always needs to go all the way to Vienna and then back to him. So because of this having a storage node on a server in Vienna will always give you a latency advantage over every other end user in the country.
Although I can only speak for Austria I’m pretty sure that almost every other country will do this similarly but probably with multiple exchange points.
Sure but we’re talking about a latency of 10-50ms typically. That is not very much compared to the latency the connection to a customer in a neighbouring country could have.
There’s of course also the latency introduced by hardware differences but ultimately I think that home users still have a very strong position against datacenters.