On Wed, Aug 21, 2002 at 12:16:20PM -0700, Dan Lockwood wrote:
Does anyone have a resource that has recommendations about how to name interfaces in a DNS name space? Is there a standard that is used? TIA
Hrm, a useful nanog discussion, will wonders never cease... Lets start by examining some examples from exiting "important networks": 0.so-5-1-0.TL2.DCA6.ALTER.NET pos1-0-622M.cr1.SFO1.gblx.net p16-7-0-0.r02.stngva01.us.bb.verio.net sl-bb22-rly-3-0.sprintlink.net ge5-1.mpr1.iad5.us.mfnx.net bbr01-p4-0.nycm01.exodus.net ges1-ge-1-1.Restonrst.cw.net so-2-0-0.mp2.Denver1.Level3.net gbr3-p40.sl9mo.ip.att.net Obviously you don't NEED to state much at all, but you probably want to come up with a naming scheme which is logical and understandable to both your engineers and the rest of the internet. The general components of a naming scheme are the geographic location, the facility information, the device information, the port information, any subint info, and optionally a speed (if you like to brag). Let's look at each one individually. Location -- Most networks use either airport codes, clli codes, or some nonstandard written-out description, each with their own advantages and disadvantages. If you are looking to describe "metro areas" moreso than specific cities, they may be for you. On the other hand, if you expect to have a wide variety of areas, clli code may be more appropriate. One of the problems with airport codes comes in defining exact boundries on overlap, for example IAD/DCA/BWI, SFO/PAO/SJC, LGA/JFK/EWR, etc. Another problem comes when the codes aren't obvious to the average person (for example, "what the heck is IAD? ORD? LGA?"). Clli codes are a little more difficult to search, but sometimes a little bit easier to figure out. Made up codes (for example CHI for Chicago, WDC for Washington DC) or written out names tend to be the most confusing. Facility information -- Most people tend to stick a number on their location code and use it to name a facility, for example IAD1, stngva01, etc. Device information -- Here is where things get a little trickier. The general idea is to come up with a descriptor for the "role" of the device, and attach a number. The fun part comes when you start trying to think up role names which are short and simple, but which people can "get" without needing some inside info or a cheat sheet. There are a number of ways you can go here, personally I'm kindof partial to GX's CR (core routers) BR (border) HR (hosting) AR (access, for cust circuits), etc. Some of the more complex ones are impossible to guess unless you know the meaning behind them. Port information -- There are a couple ways you can go here too, depending on the devices you're using. Juniper's naming scheme for interfaces solves the problem for you, with Cisco you have to get a little more creative (p or pos? gi or ge? fa or fe?), and Foundry is even worse (everything is called Ethernet). Usually you want to just replace /'s with -'s. And if you have any sub-ints, you should throw them in too. Speed -- This can sometimes be useful, sometimes bragging, or sometimes just funny when someone gets the math wrong. If you want to tack on a -oc48 or -2488M it won't hurt anything, but please don't do something stupid like sprint's 405xT1 to mean OC12. Put it all together in a way that suits you and your specific needs, and you've got a naming scheme. Personally I prefer using the hierarchy inherient in DNS to come up with something simple like: 0.ge-0-1-0.core1.iad1.yourcompany.net or pos4-0.cr1.asbnva01.us.yourcompany.net But if you're going to be one of the "one big word" or "lots of dashes" people, I (unfortunately) can't stop you. Some very good examples of a logical layout you could model from are UU/GX, and Verio. My award for most annoying layout goes to CW. -- Richard A Steenbergen <ras@e-gerbil.net> http://www.e-gerbil.net/ras PGP Key ID: 0x138EA177 (67 29 D7 BC E8 18 3E DA B2 46 B3 D8 14 36 FE B6)