January 28, 2013
A data center is the domain of several different work groups in an enterprise. A facilities group is responsible for such physical elements as the cooling plant and the power plant. The IT infrastructure group is responsible for the networking resources - everything from the structured cabling, to the networks, to the SANs and LANs. Then you have the IT systems group, responsible for the whole system, the servers, the storage systems, etc. In the days of on-premise data centers, these groups worked in close physical proximity with each other and with their end users.
Over the past several years, more and more enterprise data centers have become stand-alone facilities. No longer attached to the corporate headquarters building, they have increased the need for lights-out operation on the IT side. A new group of workers, data center ops, now manages day-to-day operations. They’re in charge of fulfilling requests from the various groups, racking and stacking the servers, connecting them to the power chain and to the network infrastructure.
What I have been seeing in practice is that data center ops groups tend to react to these requests in a totally manual and painfully inefficient way. They may send a technician to figure out the rack spaces available. They may send a second person to figure out the cabling and networking resources available in that cabinet and they may send a third to figure out what power is available in those cabinets.
Some of our customers have told me that this process can take so long - often several weeks - that by the time they’ve figured out these resources and their upstream dependencies, the rack space is gone. Lacking the tools to communicate data center resources properly among all requesting groups, the ops end up resorting to manual processes.
Good Data Center Infrastructure Management (DCIM) tools address both pain points here the repetitive, on-premise checking for data center assets, and the communication gap between departments and domains that lets those resources run out before requests can be filled.
DCIM systems start with a thorough configuration of all the assets in a data center, including their connections, and their upstream and downstream dependencies. We look at not only how much power there is, but how many network switches, ports, and power strips you have and how much power you have available in every rack - how everything is connected. You might like to know, for example, how an infrastructure element such as a circuit breaker or a network switch port may have an impact upstream on the servers that are connected to that breaker. Or you might want to know what applications are running on these servers and what departments or business units are being supported by these applications. These problems become very difficult to solve, especially on the fly and under the stress of resolving a problem.
Visualization tools within DCIM should give data center operators and managers top views of the data center, from which they can drill down to a rack, to every device and server in the rack, and even see its front and back so they can understand what ports are configured on that device.
The discipline part of the solution enforceable through DCIM tools is a consistent application of change management processes, so that data center changes and installations are recorded and the overall picture is kept accurate.
A true DCIM, then, is a holistic system with components to serve many disparate groups, whether they are the technicians, the administrators, the managers or even the VPs that run each of these groups.