Bandwidth utilization comparison thread

got this idea, will need refinement… but base concept goes like this…
post this. check below.

the full graph will make it easier to spot trends over vast periods, and the daily fan will help with more accurate numbers to try and determine node performance across many different node setups, since stuff like cancelled uploads are logged incorrectly… we don’t have much data on how well our nodes actually are performing.

Oh and try to post from the same date, and don’t go totally over board on posting them because we might want to extract them using api on a later date… will be much easier…

alas… go crazy, but try to post from the same dates as the others, as it will give us the most accurate view of what to expect out of this little experiment.

3 Likes

One of 3 nodes in my network (active since mid of March 2020). But all show very similar stats…

It‘s in Frankfurt, Germany.

2 Likes

oh maybe we should have node age on here to… that may also be a big factor…
mine is from the March 2020 and located in Denmark on synchronous 400mbit fiber

2 Likes

Maybe also the rough location :slight_smile:

2 Likes

you do seem to get more ingress data than i do currently ofc i multiply yours by 3, however have been tinkering with gpu passthrough and crashed the server many many times… so my results may be a bit outside the norm presently.

1 Like

I think the test data resumed again. I’m getting about 1.1MB/sec ingress. There’s about 8 hours left in the UTC day and I’ve already got more traffic than yesterday. Also, my dashboard with v1.6.4 is not showing the little dot when I hover my mouse over the graph. I wish this Graph either used bars or straight line segments. The curves might look nice but are not supported by actual data in my opinion. It’s hard to tell at a glance where one day stops and another begins when looking at the continuous curvy line
Node age: unknown, maybe about 1 year
Location:USA east coast
Edit: adding picture for july 3rd:

partial 4th, still in progress

2 Likes

yeah we cannot really use half days for much… it was why i picked the 3rd it was the only full day i had semi accurate data myself because i have been having some stability issues.

the graph is a bit more useful later in the month… but yeah not the easiest to gauge at…
just a few lines across it and out from the dates would help a ton.
going to make a suggestion about that…

1 Like

ok, added pic for the 3rd.

2 Likes

look how close the ingress on yours compared to mine… and i know i had like 40min to 1hr of dt due to reboots… which puts us at so close to having the exact same ingress, its almost clockwork accuracy…

doesn’t explain why striker43 seems to get significantly more data than our singular node’s…
if like @striker43 says his nodes get 3 times the ingress… that puts him at a total of 15.39gb while we are at like 11gb… tho according to storj there shouldn’t be such a big difference… so it might just be a bit random… we will see if it’s a trend when more multiple node SNO add data.

1 Like

Maybe we can create a small excel sheet or something like that to collect that data in a tabular form. Then we could add all nodes of each person who would like to contribute and get data for several days to compare. What do you think?

2 Likes

sounds good, anyone keen on doing a google drive spreadsheet and sharing it ?

2 Likes

would be nice if we could pull the data out of the webdash api and “paste” it into the spreadsheet…

i think it will be a bit more than required… from what i’ve seen and why i started this was kinda to prove a point that there might not be any difference at all, atleast between single nodes…

so the real plan was just to confirm if there is any difference between singular nodes vs multiple nodes, on the same subnet… maybe see if more nodes is better than fewer nodes…

according to the storjlings there shouldn’t be any difference… so really it might not take to long to confirm that… ofc stuff like this tend to take on a life on it’s own :smiley:

so go crazy i’m all for more data to do the statistics with… but it might not be very interesting as we progress in our collection of the information.

1 Like

Node 1:
Age: 11 months
Location: Midwest Region, USA

Node 2:
Age: 4 months
Location: same as Node 1 (same network connection)

2 Likes

Again, very close when the two are summed together. I should add that my node reached a storage limit on July 1st so my graph shows less traffic for that day. I increased the limit at some point after that. I believe my other days are more accurate.

1 Like

yeah striker seems to be the only deviation thus far…
i don’t suspect we will find much… tho one thing we should be able to see, is how much upload successrate actually matters… however the initial spark that set this idea into motion eventually was when i compared a 3 week graph between my node having 60% upload successrate and one node with 15% … and we both had nearly exactly the same graph shape and values…

so i don’t expect there to be much affect from that… it is funny how accurate the ingress is…
there can be no doubt the satellites have a very strict way of distributing data across nodes, so every node gets an even share of the daily ingress…

1 Like

04 July 2020 Final numbers


This second picture shows a sudden increase in traffic half way through the day. Did somebody flip the test data switch or does storj have a really big customer?
data

1 Like

My server still has stability issues, so the reboots i expect is why i’m slightly behind, since each restart takes about 15-20 minutes.

one thing i have been finding kinda interesting, is how low trash is on my node…ofc only had the ability to view it for a little while… but it seems really low… kinda makes me wonder if thats because my system has plenty of IOPS to clean while running the node…

or maybe its just that the node cleans every time it restarts after a crash which happens multiple times a day currently lol never something so bad it’s not good for something lol

the node data is like last check… i don’t think we should switch days to much… it’s nice to be able to compare across many nodes… ofc having a better data point than what the 3rd this month was would be good because with low data ingress also comes less detail in our comparison.

but yeah basically it seems single nodes and maybe multiple nodes are the same in ingress, thats atleast what was claimed by developers, and what the data seems to suggest and what would make sense to keep it fair for all SNO’s

1 Like

2 Likes

SGC, what version of storagenode are you running? I’m trying to figure out why some peoples dashboards look slightly different

1 Like

v1.6.4 using docker on linux
i think the windows dash looks slightly different… like total disk space has a date joined on it… also

oh and the screen capture is from chrome, not sure what version of chrome… mainly use it for local sites, so don’t really update it much…
chrome version Version 83.0.4103.116 (Official Build) (64-bit)
seems up to date… but i would think the browser might has just as much to do with how it looks, like node version has…

i know of many cases of sites that looks slightly different in different browsers, or simply doesn’t work at all :smiley: ofc most try to fix that… quickly

but they can only test for some may browsers at one time… and i often run 3-5 different browsers at once… just helps my work flow at the moment lol

1 Like