The summary is this: we would need the following number of PUBLIC subnets (for the whole campus):
/26 141 /25 10 and /26 94 PRIVATE or IFZ'd ones.Obviously, this is a very rough guess, based upon slightly broken data from CANDO.
There are of course, some discrepancies with regards to the types of devices (like VAX, VME, IOC's, AV etc) which would most likely fit into DESKTOP or something. I've basically accounted these as their own subnets for now - so the numbers above should probably be taken as an upper bound. Also there is a classification for SERVER and NULL, but it is assumed that they would need public access.
The numbers also include subnets like CAD, VISITOR and HVAC which are considered unique on each switch also. this of course also extends to the various vlans we have (ie we may get rid of certain existing subnets/vlans altogether), so the lower bound is likely to be that each building only needs one PUBLIC subnet for desktops and one PRIVATE for printers.
I've also filtered out some obvious devices from the list like bbox, farm etc. but i wasn't sure about some other switches like ir switches which are still on the list.
Please feel free to check my work. a dump of the host list (basically from cando) is here:
/afs/slac.stanford.edu/u/sf/ytl/Work/subnets/host_list.txtThe main script which does all the calculations can be run like this:
$ pwd /u/sf/ytl/Work/subnets $ cat host_list.txt | perl organise.plThis creates a summary of vlan/device/network groupings. it also creates a summary of the required number of subnet sizes for each switch based upon whether it is ifz or public. The output of this is currently stored at:
/afs/slac.stanford.edu/www/grp/scs/net/racks/slaconly/number_of_hosts_in_each_vlan.txtThen I simply do a count on the number of each size of subnet.