I have full duplex Gigabit internet, my server is on ethernet and using SSD.
In order to server clients with fast connection and demands for speed, the way i can try to do with my node settings would be to increase storage2.min-client-upload-speed and limits the storage2.max-concurrent-requests.
This would mean i would not get be able to serve as many clients as i would like (due to current apparent lack of ability on the node to separately configure the ingress vs egress capabilities or other configurations which maybe are there and i did not yet identify).
But would allow me, for the clients that i do accept and who will come back to me later to get their data, to serve them at 5ms latency to london or germany, which i have and with SSD speed limited only by Gigabit.
My one concern is, by increasing storage2.min-client-upload-speed i would be forced to refuse slow clients in order to preserve my “hot” data capabilities for clients with the speed and need for hot data access patterns, even as limited as they are.
But I wanted to ask, is there any reputational risk to my node if i do this storage2.min-client-upload-speed to say 5Mbit or more, so that i keep my 10-20 “slots” (concrreunt requests) available for faster clients when they want to upload/download?
This will allow me to remain within 5ms latency and offer a large portion of my internet bandwith for those high end users.
Going with what alpharabbit said: you may be vastly overestimating the amount of traffic you’ll receive (unless they start uploading test data again).
Like this guy has 130’ish nodes. And if we look at his Weekly traffic graphs the average ingress is 350’ish Mbps, and egress is 150’ish Mbps. If you squint at it a bit… that’s around ±2.7Mbps inbound and ±1Mbps outbound averaged per node. And he’s on a 10G connection.
So… tuning your 1000Mbps connection may not add much
True, things will burst higher. And true, we’re around three months into a low-traffic period. So thing will hopefully pick up soon…
(Also: I just noticed you made a couple performance-tuning posts recently: so remember any answers here aren’t meant to be discouraging. It’s just other SNOs saying what traffic they’ve been seeing)
Please, disable this setting, it hurts customers. The default value (0 - no limit) is a best for almost any hardware, except so called “potato” nodes, i.e. power efficient SoC devices like Raspberry Pi. These tiny things may not survive under the high load, so you could enable this limit. It also can be used if your HDD is SMR, because these type of disks are not suitable for the node (or anything in my opinion), so you could be forced to use the limit otherwise it will stall or even crash.
In any other cases it should be disabled (commented out or have a 0 value).
Please note - the usage depends on the customers, not hardware or software settings. Changing these settings may result in worse performance, so I would like to suggest to do not change them.
There’s a saying… you can’t see the forest for the trees.
Any company should portray success as best it can (with a #hitload of disclaimers, of course - lol.)
2 cents,
Julio
P.S. As of Dec 31st, I think they will have finally purged everything they can, or needed to from the network.
