I have a group I'm assisting in setting up a unique network. It's a mining operation - but my only interest/control is in the networking.
Up to 800 miners. There is no need to speak to other clients on the LAN other than the gateway, and max bandwidth per client is about .3Mb/s
Network from the outside in:
2 wan devices (ATT/Comcast)
1 4 interface pfsense device (2 Wan, 1 Top of Rack Switch 1, 1 Top of Rack Switch 2 - Hopefully in 802.1AX)
2 48 Port top of the rack switches stacked via 10Gb sfp+
~40 24 Port Gigabit distribution switches each with a home run back to each top of the rack switch (2 total top of the rack switch connections - hopefully in 802.1AX/AD)
~20 miners per distribution switch.
In this instance, the single points of failure are the pfsense box (which will have a hot spare, we considered a CARP or similar but dismissed it based on how simple it would be to swap out the 5 cables) and a single distribution switch which would affect up to 20 devices (there will also be a spare switch available).
I'm hopeful I can find top of the rack switches that will stack via 10Gb and allow 802.1AX across difference units. Short of that, I'll enable spanning tree and just let the switches "work it out". The goal here is to prevent failure of a top of the rack switch from killing 1/2 of the operation.
Is there a switch that allows LAG across units of a stack and that doesn't have a limit on the amount of LAGs created (other than physical ports? Will the Ubiquiti US-48 do it?
Anything I'm missing?
Edit one: fix transposed label in map
Up to 800 miners. There is no need to speak to other clients on the LAN other than the gateway, and max bandwidth per client is about .3Mb/s
Network from the outside in:
2 wan devices (ATT/Comcast)
1 4 interface pfsense device (2 Wan, 1 Top of Rack Switch 1, 1 Top of Rack Switch 2 - Hopefully in 802.1AX)
2 48 Port top of the rack switches stacked via 10Gb sfp+
~40 24 Port Gigabit distribution switches each with a home run back to each top of the rack switch (2 total top of the rack switch connections - hopefully in 802.1AX/AD)
~20 miners per distribution switch.
In this instance, the single points of failure are the pfsense box (which will have a hot spare, we considered a CARP or similar but dismissed it based on how simple it would be to swap out the 5 cables) and a single distribution switch which would affect up to 20 devices (there will also be a spare switch available).
I'm hopeful I can find top of the rack switches that will stack via 10Gb and allow 802.1AX across difference units. Short of that, I'll enable spanning tree and just let the switches "work it out". The goal here is to prevent failure of a top of the rack switch from killing 1/2 of the operation.
Is there a switch that allows LAG across units of a stack and that doesn't have a limit on the amount of LAGs created (other than physical ports? Will the Ubiquiti US-48 do it?
Anything I'm missing?
Edit one: fix transposed label in map
Last edited: