Storj updates impact bandwidth utilization

Some node updates (or node restarts? or both?) appear to have a clear and significant impact on my bandwidth utilization.

Note v1.6.5 (“No code changes. We want to force a storage node restart in order to observe the impact on the satellite side.”) released 7/20:

This dip typically precedes a long, slow, painful climb back up from that baseline over the next several weeks, which you can see here. Note though that v1.11.1 released about 8/25, did not appear to have an impact:
(Second cool graph goes here, which shows consistent steady progress/traffic throughout August. Too bad new users can only include one image.)

…until it happens again. Most recently, on or about 9/9, when v1.12.3 was released:
(Third not cool graph goes here, showing another dramatic drop-off from 100-135GB daily to 40-60GB daily, beginning on about 9/9. Too bad new users can only include one image.)

1Gb/s up/down internet connection. Running on Unraid 6.8.3, Docker, and oreandawe’s Storj v3 automated docker build. I use the CA Auto Update Applications to automatically upgrade all docks four days after initial availability.

I’m under the impression that software updates should not impact reputation. Node is currently at, and tends to maintain, Suspension and Audit scores of 100% on all satellites (minus occasional brief downtime for hardware upgrades). This node has close to 12TB used on it, so my gut says that at least downloads should be fairly consistent week over week, unless activity across the network is still -really- variable? These significant drop-offs just seem odd.

Can someone help me understand why I’m seeing these drops in bandwidth utilization? As one can imagine, this is having a significant impact on sustained income over time. Thank you in advance for any insight, guidance, or advice you may have.

We can also find a significant correlation between weather and traffic. Maybe the customers download less while they are sitting in the sun? More download traffic on a rainy day?

Makes no sense right? Why do you think that a node restart would change the customer behavior?

4 Likes

Thanks for your reply.

It’s not that I think that a node restart would change customer behavior, but rather that a perceived change in reputation results in satellites favoring other viable nodes. Yes it may just be customer traffic flows, but if so what’s odd to me is that the traffic steadily increases over time before collapsing suddenly to a baseline. Are others seeing that pattern? Do satellites consider any other factors besides Suspension and Audit scores?

Not yet. The nodes selection is random

1 Like

Just to ensure that I understand, please confirm:
Each and every time any upload/download transaction is requested, a node is selected at random from those nodes available? Yes?

That does seem like the right place to start, though my graphs still remain intriguing. Perhaps indeed it is just a pattern in customer(s) usage that is not apparent on the surface.

Thank you Alexey.

This is my traffic graph for comparison


There is some kind of pattern

2 Likes

It is the weather for sure :smiley:

In my case I am running a backup with duplicati from time to time. Even if I pick a random day to execute the backup job it will pick up more files on a weekend just because that is the moment I have time to work on my tax report and other paper work.

5 Likes

Sometimes there is a clear pattern depending on the time of day.
Number of concurrent Steam users for example:

Though I guess there would have to be more Tardigrade customers for such patterns to emerge.

Lower uploads after a node update might also occur if you have a slow HDD and not all uploads can be processed after a restart of the node due to it checking all the files keeping your HDD busy.

2 Likes