Do mind that the ANT_MAX_STREAM_DATA setting is currently in main, so not released with any binary currently (as far as I’m aware). You could test if the value can be tweaked by using an invalid value (ANT_MAX_STREAM_DATA=foo) and you should find QUIC connection receive window value override failed in the logs.
Currently we upload chunks in parallel. The number depends on the system and is logged as Chunk upload batch size: <value>. It roughly corresponds to CPU cores * 8, but CPU cores mean different things on different (modern) systems.
I just skimmed through this topic as I was mentioned. If there’s anything else people like to know from the current codebase, let me know.
Just saw this and could be the reason why some cannot upload from their home connections but can from a VPS data centre one. For the reason the topic explores.
Yes, this is definitely something to look at. I have quite a good CPU and available_parallelism yields 24, meaning CHUNK_UPLOAD_BATCH_SIZE=192. But that doesn’t say anything about my internet connection.
was about to ask @Southside if he self-built the binary to include those changes … because I couldn’t find ANT_MAX_STREAM_DATA in the source of the current release …
which means this is just usual upload variance for the current network
I think for these tests ( @Southside too and thanks man) it’s best to test downloads and not uploads. The upload payments etc. will screw results, but downloads is a more direct measure of network performance as we don’t have the payment steps which will always be slower as it’s more negotiation.
@bzee@dirvine While responding to another question I realise another major benefit of reducing the max window size and it isn’t for the router as much as a good number of people still on sub 40Mbps upload speeds.
One full chunk is 32Mbits of data to be sent. This represents a significant amount of time for the chunk to monopolise the uplink. At 10Mbps uplink its 3.2 seconds worth and in basically one block of data. 40Mbps it is 0.8 seconds.
With a greatly reduced max window size then the (at least the initial) chunk upload (attempt) will be 128KB or 256KB (1 or 2Mbits) before causing a gap in transmission allowing other packets queued up to be sent to the internet resulting in better experience for other people using that internet connection.
This is perhaps as important as the router buffer overflow issue.
And another reason why a relatively small max window size is important for helping home users run nodes with a large max chunk size
We are after all wanting as many home users as we can possibly get and with the opportunity now to run even more nodes on a given computer (EDIT: thus more uploads of chunks happening) the internet experience for other users on that internet connection is very important to rates of adoption and not have people give up due to their internet crapping out every time a block is sent to another node. Even @riddim Experiences this at home from the few reports he has given
Yes I’m at 0 home nodes - not possible to run nodes if that makes me freeze every now and then in teams meetings for a moment and missing a word here and there when working from home…
The intermittent video streaming in the evening is annoying (and most certainly enough to give up home nodes for 99% - including myself Medium to long term …) … But not being able to run video calls in parallel is an absolute killer
(just to specify a bit more; I’m having a 50/20 MBit internet connection; which on paper should allow a smallish number of nodes … but those internet-blackouts that kill my connection on every chunk that is passing by don’t allow me to run local nodes … it was different before the upgrade that included [many changes and I shouldn’t focus on that singled one …] the switch to 4MB chunks)
and btw - in my router I already set priority for traffic of my local node runner to be lowest of all machines and raised priority for others … but didn’t resolve the issue; I’m already not using the ISP supplied router but a Fritzbox (which is a couple of years old now …still better than the thing I got from my ISP …) and still available settings don’t seem to be enough to get my local nodes running without interfering with my work …
Fwiw, Zen Internet (UK) are my provider and they dish out FritzBox 7530s. They are a premium UK broadband provider.
I switched out the FritzBox for a MikroTik and it was much better. The FritzBox couldn’t handle the connections for ant clients or ant nodes. It was hard to upload anything (a few versions ago at least).
yeah I know and was thinking about switching the router … but “switch out your router agains one that most people will have a hard time to configure right; then you can use your preexisting hardware” doesn’t sound the same as “run what you brung” … right?
… and we can say that’s needed for node running and API usage of autonomi … but I was hoping this would change at some point …
