Are you sure? 
Global stats are not so positive
one or few customers´ testing should not necessarily change stats on a global scale
Christ, I must be losing a LOT of races!
Uploads (ingress) or downloads (egress)?
I’m definitely not complaining about the egress. It’s quite nice to see. But is it not a little concerning that these scalability tests are saturating even gbit+ connections on the node end? Especially given the stated intention to keep node numbers relatively low for now? It’s a problem you could easily scale out of with more nodes, but if a single (or few) customer can saturate nearly all node connections, that doesn’t sound ideal.
2 Likes
If hot files would have been implemented already, then the files in question would be distributed on more nodes.
Sounds like exactly the case why hot files are required.
It seems that everyone is already experiencing these spikes though. So the files seem to already be distributed plenty across all nodes.
I don’t know if ‘everyone’ applies here as I believe only a small percentage of node operators are visiting the forum and voicing their thoughts.
And we don’t know exactly yet (we may never know) who this customer is and what exactly they are doing. I can assure you that I have still plenty of space and bandwidth available. They can put much more load on us if you ask me. 
1 Like
Out of 3 nodes on two distinct subnets I only had spikes on 1 node.
2 Likes
Interesting, I saw it across all my nodes, except ones that have been full for a long time.
Yes, my oldest node that was largely full except for the test sat deletes didn’t have a spike.
I would imagine that for every one 1Gbps connection that is saturated there will be plenty more (like mine) who are barely touching 200mbit
Once all the big boys are saturated us small fish will start winning more races. I suppose it might increase latency somewhat but doubt bandwidth will be grossly affected.
2 Likes
Interestingly my older nodes seem to have far fewer spikes. I guess it must be more recent data.
I am wondering if this could also have some impact:
I’ve been noticing those spikes since mid-August.
The peaks have been much lower than other operators. The day I had the most traffic was August 19 with 107Gb upload.
I think if we get to a point where all the available nodes are saturated and spiked on Egress bandwidth, we will all be visiting the Lambo thread and thinking about making appointments for test drives.
I understand the limited number of nodes with pieces issue, no need to retort. The Storj engineers will adjust the software as needed to accommodate potential bottlenecks. There may be an opportunity for Storj Labs to make additional revenue if customers need a higher rate of availability than what the current protocol supports. Since they would potentially need more pieces on more nodes, there would be additional costs related to storing those pieces on more nodes. So, this “problem” could be a good thing for the business as a whole.
5 Likes
Sadly, many of us in NA are stuck with abysmal upload speeds. I can still get 100GB+ egress on some days, but that’s almost 1/4 of my maximum bandwidth.
My old strategy is to fill a 14Tb hard drive.
The hard drive didn’t seem to respond well with 10TB full.
My strategy has changed. I am emptying the node to change it to an 8Tb disk. Now I will have two nodes on two 8Tb disks. By dividing it into two disks I hope to have a little more capacity to serve the data.
My connection is 600Mbps up and down. Could downgrading to a 300Mbps up and down connection have any impact or limitation?
The 300Mbps connection is 3 euros cheaper every month.