New Subnets 2008

Les Cottrell & Yee-Ting Li, created 8/23/08, last updated 9/1/08

Introduction

Due to security, network performance and network resilence requirements of next generation local area networks at SLAC, all hardwired, non-DHCP hosts (visitor) will need to be re-IP'd over the next few months.

Each building will be required to have it's own subnet(s) and there will no longer be large spanning tree's that stretch across campus. A prime example is that of the PUB-'n' subnets.

This document provides guidelines on the both the subnet and vlan naming and numbering procedures.

Location Naming Convention

All official SLAC buildings on campus are designated an unique three digit number. This number, appended with sufficient other information shall be unique and help to quickly identify a specific location on campus.

The physical building location of the device shall be included in the form 'Bxxx' where 'xxx' indicates the numerical label of the building. Names and aliases shall not be used. * problem is that there are actually 4 digits....
If the building has many floors, then the floor on which the switch is kept shall be appended onto the building number in the form 'Fn' where 'n' is the floor number in the American convention (floor 1 is the ground floor). NOTE: Basements?
Should there be many racks on that building, the rack identifier shall be used instead of the floor (as typically the rack number shall include floor information). The rack id shall be prepended with 'R' to indicate the rack.

Examples:

B115
B044f1
B050R2AH08

Network Device Naming Convention For Buildings

Due to the number of buildings present on the SLAC campus, a flexible naming scheme to quickly identify the location of switches was paramount. It was decided that a functional approach to switch naming would be inflexible due to the dynamic workforce at SLAC and the ever changing roles buildings take.

The type of device will be preprended onto the name. Current valid types of devices are Layer 2 Switches: 'SWH-', Routers: 'RTR-', Hubs: 'TT-', Terminal Servers: 'TS-'
The location of the network device shall then be appended. The naming conventions are outlined above.

Examples:

  swh-b050f1
  swh-cgb2AH08 *** haha.. doesn't work...
  rtr-kavli1 *** nor does this

Subnet Naming Convention

The purpose of subnet naming is to provide transparency and quick identification of the location and function of hosts contained within that subnet. As such, the following convention should be followed:

The general format is to utilise the switch naming convention to quickly identify where a subnet terminates and to also include the type of hosts connected to that device. The latter shoudl provide us flexibily to further segregate by functions should we choose to expland those in the future.

The subnet name shall begin location of a device as outlined under 'Location Naming Convention'
An enumerated list of host group types shall be appended to the end of the location. This list is TBD.

Examples:

B280-PUBLIC
B044f1-IFZ
B050-TRAINING

VLAN Naming and Numbering Convention

Unqiue VLAN numbers are not really necessary as we shall not be spanning vlans beyond the connection between the BLDGCORE and individual building switches. However, in order to reduce potential problems with trunking the wrong vlans etc., unqiue vlan numbers are recommended.

Simarily, to aid with the human interaction, the names of those vlans should also be unqiue.

The name of the VLAN shall be identical to the name of the subnet that is carried upon it. This assumes a one-to-one mapping of vlan to subnet.
VLAN number shall be allocated depending upon the IP and mask of the subnet that is carried upon it. It will also use the SLAC vlan numbering scheme.*** More info on algorithm etc.

Subnet Ranges

The minimum size of a subnet will be 64 (i.e. a /26, see Subnet masks). In general subnets will be sized to be at least twice the number of existing hosts in the subnet. N.b. some addresses are reserved in each subnet (e.g. 4th quadrant for /22s are currently used for IFZ, 1-15 are used for darknet scanning - this number of hosts for darknet may be reduced).
PUBLIC subnets numbers will be chosen from the free subnets in our public address space (class B 134.79.xxx.yyy). We will also look for subnets that we can reclaim (e.g. no host pingable and no host in CANDO, the latter hosts are blocked anyway ), see for example reclaimable subnets and switch hosts. Jared volunteered to contact the owners to start this process.
Printers and IFZ hosts will go in class B PRIVATE subnets in the 172. range, one subnet for each group per switch. IFZ-LITE is not being implemented until we get the equipment/resources.
We will try and allocate the PUBLIC subnets in contiguous ranges to simplify firewall ACLs. We may also try and leave empty /26 subnets available next to used /26 subnet to simplify later expansion. There is also consideration to keeping big chunks available where possible.
For the old scheme for /22 subnets the router is the first address in the fourth octet. For /24 subnets the fourth octet is 1, for /25 and /26 subnets the router is the first address in the subnet range.