When you climb into any car, you will find a set of common elements. While placement may vary slightly, overall...
the components do what we expect them to. The fuel gauge tells us how much source of motive power remains; the speedometer relates our velocity of motion relative to stationary objects; the brake pedal initiates stopping; and the big round wheel in front of you allows you to change or maintain the direction in which you wish to go. From Corolla to Cadillac to Carrera, the dashboard remains basically the same.
Step into a data center, however, and there is no such dashboard, least of all any commonality from one monitoring system to another. Despite the fact that every data center shares the same basic components -- backup generators, cooling systems, Ethernet networks and uninterruptable power supplies (UPSes) -- monitoring systems offer zero commonality.
Instead, each vendor's product has some proprietary management interface with no common data access methodology. It's as if Toyota decided to put its dashboard in the trunk and BMW puts its dashboard under the hood. That would make no sense and we shouldn't accept this in data center monitoring systems anymore.
The recent Data Center Pulse gathering produced a Top Ten list of data center user demands. One stand-out item from that list addresses this very issue: Users want a common communication standard to monitor all layers of the power delivery system, connecting building management and IT systems.
Vendors will be resistant at first, as they think (as they always have) of management software as a lock-in or a value-add -- a way to sell their gear over somebody else's. They hate "open" systems, which they view as uncompetitive.
But the vendors ignore the lessons learned by every other industry, where commonality of the interface actually frees them to concentrate on the critical functionality of their systems.
Those of us who actually use the equipment know that there are two purposes of management software: setup and monitoring. Setup is vendor critical, and should be left to the vendor to control every aspect.
Monitoring, however, is frequently the collection of mundane data -- the basic dashboard concept. The queries are simple and the answers are even simpler. Questions such as: Are you up? How many total bits, kVA, etc. are you using? How many bits, amps, etc. are being used on that specific port?
The answers to those queries are very important to a network operations center or facilities manager, but mundane to the vendor. They are the data points that when viewed through a management console are like the dials and gauges of a car's dashboard.
Data center managers don't want to have to look under the hood all the time to see how the machine's components are working, they just want to glance at that dashboard. Every vendor wants to sell you a specific dashboard along with its widget, be it an HVAC unit, a UPS or whatever. But you only need one dashboard for the whole facility, not one per major subsystem. So what's needed is that common communications standard that Data Center Pulse's list demands.
A common communications standard actually already exists and is used to monitor several parts of the data center. Namely SNMP, or Simple Network Management Protocol, which has been around for almost 20 years. It works. It's scalable, reasonably secure and, best of all, extensible. It does that mundane status monitoring task very well. Counting bits is no more complex or easy as counting kVA or BTUs, so why not extend SNMP into data center infrastructure? In reality, if every device supported SNMP, the current dashboards we use on the IT side could be extended to cover them.
This battle has been fought before. In the emerging networking market of the late 1980s and early 1990s, every vendor had a different management software package for its equipment, and while some of them were very good, none of them interoperated. Users were frustrated by this situation and demanded support for a universal standard -- SNMP -- from network equipment vendors. Today, SNMP support is nearly universal and it hasn't hurt the networking industry one bit. We need to introduce data center infrastructure vendors to SNMP and RFC1157.
Suddenly all of those massively useful, well-established SNMP-based tools that are already monitoring the network can be used to monitor the entire facility. The dashboard can and will provide us with at-a-glance knowledge of power consumption, cooling capacity, network usage and the status of the backup power systems.
No wheels need to be reinvented, just a single, simple, open standard adhered to across vendors. Next time you speak to an equipment vendor, ask if it supports SNMP. Once vendors are all onboard, we won't have to drive our data centers blind. The gauges and dials will be right where we expect them to be, providing us the data we need.
ABOUT THE AUTHOR: Chuck Goolsbee is an executive at colocation provider digital.forest in Seattle, Wash. He has achieved notoriety blogging about the obsolescence of raised floor in the data center and for threatening to gas server designers from Dell with FM200. In his spare time, he enjoys wrangling geeks and tuning SU Carburettors. What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at firstname.lastname@example.org.