LAN Party L3 Networks without additional Hardware
2018-05-18
- LAN Parties need Broadcast traffic for network discovery
- Many Networks rely on Spanning Tree and Fat-Links as Uplinks.
- Hub-Spoke Network Design
- Single Point of Failure
- Real topology on the floor is more like a grid of mesh
- L3 Networks can provide Failover and Load Balancing
- requires additional hardware
LAN Party networks need broadcast traffic for the local game discovery. This means that LAN party networks are usually a large layer 2 network. But large layer 2 networks with several hundret or thousand hosts are not that nice because the amount of broadcast that normal (Windows) computers send grows rapidly. Usually those networks are large hub-spoke designs. But this often results in very long cable runs if every switch has a direct to the core switch. The alternative to very long cables would be daisy chaining a lot of switches (or even building loops which would be disabled by spanning tree)
The "L3" networks I have seen are still a hub-spoke design on L1 and L2. Each access switch puts all its access ports into a VLAN. The gateway of that VLAN is somewhere on a core switch which does inter-VLAN routing. To get the required broadcast traffic between the seperate VLAN all those VLANs are also connected to a server. This server then uses software like [2] or [3] which listens for broadcast on all those VLANs and duplicates the received packets on all other VLANs. Depending on the software that is used either a white- or blacklisting is used.
This design has several disadvantages:
- The server that bridges the broadcasts is a single point of failure.
- The core switch is a single point of failure for the whole network
- The network cannot use redundant links for traffic because it still has to use spanning tree.
1 [2](Link zu service-discovery-helper) [3](Link zu bcast-bridge)