Tag Archives: FNN

To follow up the previous post, the group finished wiring HAK Friday afternoon… and into the evening. About 480 runs of Cat5 for both the FNN and a separate network for monitoring and provisioning. It’s an interesting looking network; because of the way it was designed, it is unusually symmetrical for an FNN, so the core has trunks for the rack-crossings on each network, and is therefore rather neat:

Unfortunately, this does mean that all the cables in each wing are passing through the small slot in the center… which looks about like you would expect from the back:

It should produce some interesting results.

I’ve spent a good fraction of the last several days helping with the preparation and assembly for a cluster the research group is building from 96(+4) old Athlon XP nodes to conduct network research on. The machine is named HAK for “Half-powered Athlon cluster in Kentucky, referring to the fact that every pair of nodes is sharing a power supply (notice the flipped cases). This design is both to save power (switching power supplies are most efficient when heavily loaded), and save power supplies, since that is the part that has had the highest failure rate in the pool of machines used to build it. After a couple passes of repairing nodes, homogenizing network cards and the like, we got to the photogenic part today: attaching the network. In the picture, there is a standard tree topology 100Megabit Ethernet network across all the nodes (the yellow and pink wiring), which will be used for administration and provisioning, and began sticking a full FNN (the bundles of red cable snaking around… those are the first 3 sets of 32, the other 9 are the colored bundles on the floor). That machine is set up to have it’s network swapped out with a FFNN (Fractional Flat Neighborhood Network) and SFNN (Sparse Flat Neighborhood network), but requires this initial fully populated Universal FNN as a baseline for comparison.