Template:Unicast vs. Multicast vs. Torrent-Cast images
From FOG Project
Unicast
- Accessible from Host Management --> Select Host --> Basic Tasks
- Accessible from Group Management --> Select Group --> Basic Tasks
- Accessible from Fog iPXE Menu
- This occurs when scheduling a "Download" task, "Quick Image", or during "Full Registration" the user selects image after registration. The server (or storage node) will directly send packets (the image) to that Host's MAC address.(aka. TCP) This is excellent for directing network traffic. This can be performed under Groups running multiple unicasts to multiple hosts. However, all the hosts will not finish at exactly the same time.
- Since unicast does not rely or communicate with others in a group it is unaffected by "hung up" host
Multicast
- Accessible from Group Management --> Select Group --> Basic Tasks
- Accessible Image Management --> Multicast Image
- This occurs when scheduling a "Multi-Cast" task from Groups. The server (or storage node) will wait for all Host computers to be network booted. Once all hosts have registered for the task the packets will be sent out out to the entire network.(aka. UDP) Packets are sent and received exactly the same. Image may not complete exactly the same time
- Down side to this is if one host gets hung up then the entire process is stopped till that host catches up
- Multicasting
Torrent-Cast
- Currently not working.
- Accessible from Host Management --> Select Host --> Basic Tasks --> Advanced
- Accessible from Group Management --> Select Group --> Basic Tasks --> Advanced
- The image is broken down into chunks and indexed to create a torrent file.
- The server (or storage node) beings creates the torrent and starts seeding. The hosts request the torrent and start downloading chunks for the torrent. Once a chunk is received then the hosts will seed that chunk to other hosts. Once torrent is done downloading it will start imaging. Then the partition is rewritten to accomidate the size needed to download the files originally.
- If one host gets "hung up" the other host are not affected, but marginal time and speed are lost.