This morning I changed the configuration of the EPICS gateway processes that serve the vacuum subnet IOC PVs to limit the allowed PVs to those actually available from the vacuum IOCs.  This is to eliminate the additional broadcast traffic that occurs when the gateway searches for PVs that do not exist in this subnet (and it does this a lot, each time a client broadcasts for a non-vacuum channel).  Searches are now limited to channels starting with "HVE-" or "H0:FMC-" (this last should probably be constrained further, but this is a good start).  Attached is a plot of the reduction in broadcast traffic as a result; the 'mini-peaks' that still occur near the start of the hour are probably from the hourly burt backup for H0:FMC channels not in the vacuum subnet.