To DNS or not to DNS: That is a question!

Here at UMHS, some of our network devices, i.e., switches and routers are named in some sub-domains in the DNS system, which provides an easy way for people to access them on the command line. For example, it is much easier to remember some-clinic.switch.med.umich.edu than 192.168.129.18.

To name our devices in a consistent way across the board, we have to decide on a device naming convention -- That is likely a common practice for all IT teams out there managing a network of any size larger than one single building. However, that presented me with this issue: Do we continue with naming the devices in DNS or not?

There are multiple ways a device in the IP networking world may be identified, DNS name is one of them, IP address(es) is another. Yet another is the SNMP MIB object SNMPv2-MIB::sysName. They each serves a different purpose: a DNS name serves humans and an IP address serves machines. The sysName object, on the other hand, serves network management systems that talk to devices using the SNMP protocol.

To minimize conflict of interests in naming a device, it is probably the easiest to make a DNS name identical to the sysName on the device. However, I have seen in more than one organizations where device naming conventions factor location and IP address information into the sysName, which in Cisco IOS is configured using the hostname command. Adopting such a naming convention effectively renders the human readability of a DNS name close to that of IP addresses. Also, synchronizing the sysName values defined in devices and their DNS names take a lot of work to do right. Some times it is a task crossing team boundaries, which is not to say it is impossible but difficult to say the least. So many a time the DNS names are out of sync with the IP addresses of devices after they are upgraded or a network is renumbered. The Combination of those two factors probably explains why I have seen many of my colleagues access devices using their IP addresses directly rather than using their DNS names.

My thoughts are that, a device should be named just for human access. Therefore embedding location and IP address in device name may seem to be a good idea but defeats the purpose of a name. After all, there is a sysLocation object in the SNMPv2-MIB and DNS exists to translate a name to an IP address. What we really need is probably a simple utility which, when given a DNS name or an IP address of a device, produces the other information such as location or whether the device really is a router or a switch.