* Labeling data center equipment today
In a rapidly virtualizing data center environment, what do we need to label and how do we meaningfully label things? If you are running a data center, you know that you still must label the physical objects, but how you do it will have to change with the changing face of data center operations.
Anyone who has run a data center knows the problems with cable, switch, and server labeling: it gets out of sync with reality. Many label plants, if you’ll allow the construction, were only accurate for as long as it took to get from go-live to the first change event.
Truly meaningful labels, such as “this switch port serves that server,” became very difficult to maintain even before the advent of virtual servers, as the re-purposing of servers and mobility of applications became prevalent. Less meaningful labels, such as “this switch port goes to that patch-panel port” are hardier but less helpful in rapid troubleshooting situations.
The accuracy of server labeling is also notorious for how quickly it becomes completely inaccurate or – and this may be worse – partially accurate, as what users see as IT services continue to evolve into interoperating components on many servers.
There’s good software out there for tracking physical assets, but the key weakness in all of it is that it either requires physical input by data center techs to stay up to date, or relies on automated tools for discovering port information. Such tools cannot collect any information about the passive components of the infrastructure – the cables and patch panels! Also, they cannot automatically re-label a server or a rack to reflect what they are doing this year (in a slow-changing, manually managed environment) or this week (in a fast-changing one), let alone what they are doing today (in a fast-changing, automated-provisioning environment) or even right this minute (in a fully virtualized milieu).
Yet we still retain a very real need to reach down to this most basic layer of the infrastructure and know where the traffic is really flowing, what is actually on a box on which we are seeing warning lights.
The answer might be in the creation of “semi-active” infrastructure and a new standard for “semi-active” labeling on servers and racks. Consider the utility of a patch panel infrastructure that could, on demand, switch its ports from passive to active connections, determine which ports on each were connected to which devices or which other patch panel ports, and then revert to passive mode.
All that information could be displayed on cheap, slow LCD panels. At a minimal and momentary cost in increased latency on network traffic, then, such a system could collect a complete and accurate set of physical network connection data.
If there were a standard for exchanging this information, intelligent label panels on the fronts of servers or racks – again, cheap, slow LCDs – could display information about how servers were physically connected. With a standard in place, enterprise management tools could send information about applications and services currently running on a server (in whole or in part) for display as well. Such management tools could also pull the information from the label plant, and use it to update inventories and maps of the systems.
Ultimately, then, the answer to the question “What do we need to label?” hasn’t changed; we still need to be able to map the virtual to the real. What has changed is the answer to “How do we need to label things?” As with so much else, the answer has become “Dynamically and on-demand.”
The alternative is simply accepting that, however careful our notes, our understandings of the physical infrastructure are nothing but antiquated notions. (Thanks to the brilliant Thomas Dolby for the line I used as my title.)
