From "Nice to Have" to "Need to Have": How CFD Fits Seamlessly into Data Center Operations

2 August, 2018

The phrase “garbage in, garbage out” is quite relevant for any analytical model. This principle is perhaps even more relevant in the data center industry, where use of simulation isn’t as pervasive as in the aerospace, automotive or consumer electronics sectors. One of the biggest (and most valid) criticisms of data center simulation models is that they get outdated quickly, as data center changes are happening all the time.  The reputation is that a data center model represents a snapshot in time, and thus can only be practical for conceptual design or one-off projects. If the model doesn’t keep up with the real facility, then it becomes extremely difficult for it to be utilized in operations.

Another common disconnect with simulation in operations is that many organizations do not have resources that can be specifically dedicated to managing CFD models. Facilities and engineering teams have limited bandwidth, and data center CFD modeling is seen as a "nice to have, not need to have." Not an ideal recipe for adoption.

To better meet the needs of the industry, Future Facilities has built several native API integrations that allow the Virtual Facility (VF) model to automatically update in parallel with existing monitoring systems and IT workflow tools.  Our objective is to ensure that the VF is always up to date and ready for simulation. These integrations enable 2-way data exchanges between common DCIM/ITSM solutions and 6SigmaDCX via a software module called 6SigmaGateway. 6SigmaGateway runs continuously in the background, gathering live data and planned changes which are then synchronized with VF models. 

Figure 1. Example demonstrating the flow of information from a DCIM database to the VF and back.

6SigmaGateway supports both SOAP and RESTful web services and can communicate directly with SNMP-enabled devices such as rack-mounted PDUs. The API integrations are seamless and do not require changing a single line of code from the source database. This means that the VF model can be integrated into operations without any disruption to an established workflow.  Rather, operations teams gain access to a digital twin of their facility that is both a visual database and capacity planning tool, in addition to their existing tools. 

Let explore how this works. The following step-by-step process highlights the capabilities of synching VF models with DCIM/ITSM platforms such as ServiceNow™, Nlyte®, StruxureWare™, Trellis™, and RF Code.  In these cases, we show that the VF model can also be initially built from scratch using information pulled from the database.

1) Model Outline Import

If a DCIM tool already represents the facility as a virtual model with a dimensioned footprint, these coordinates can be imported to automatically define the walls of the VF model.  This is generally a one-time operation and can be skipped in all subsequent database syncs.

2) Infrastructure Import

At specified intervals, 6SigmaGateway runs a "Discovery" for objects in a DCIM database. During this process, information is collected on objects such as cabinets, floor grilles, sensors, and power & cooling equipment. The data collected typically includes location, manufacturer, and model.  Likewise, IT equipment and blanking panels can be imported in their correct cabinet and RU positions and labeled with the proper asset IDs.

A first-time sync goes much of the way towards building a complete VF model from scratch.  Only perforated floor tiles and piping/cabling are left to be modeled since these objects are not included in the database.

Regular syncs reconcile differences between the source database and the VF model.  Only new objects are imported, while existing objects are ignored. During import, since discovered objects have a specific manufacturer and model number, the VF attempts to map these database fields with matching objects in the 6SigmaDC Library. 

Database items that do not find a match are still imported as generic placeholder objects, and the missing library items are provided by Future Facilities as part of ongoing customer support.

3) Power/Temp

6SigmaGateway also runs "Monitoring" at specified intervals for any discovered devices that report power, temperature, or humidity readings.  These devices may be environmental sensors, rack PDUs, IoT devices or servers with built-in telemetry.  Over time, the monitored data points will create a record of power utilization and environmental conditions.  When running a sync, the user has the option of importing either the most recent, peak or average monitored data points over the desired date range. 

Figure 2. Example showing monitored rack PDU data in 6SigmaGateway that has been synced to a VF model

4) Sync Data Back

Once a VF model is fully synced, it is ready to be simulated.  The CFD simulation calculates the distribution of temperature and airflow in the room, based on cooling control set points and monitored power utilization.  As a result, the model quantifies the temperature and volume of cooling airflow (CFM) delivered to each IT device in the data center. These calculated CFM values are directly translated to a kW cooling capacity for each cabinet. This cooling capacity is passed back to the DCIM solution, which in turn updates the database's cabinet capacity limits.

With a pinpoint understanding of cooling availability, data center operators attain a complete view into the what, where, and why of potential capacity bottlenecks.  This allows them to stay a step ahead in their capacity plans and better optimize IT deployments. And it can be accomplished without drastically altering existing workflows or dedicating significant resources to model management.  

Blog written by: Aitor Zabalegui, Lead Consultant Engineer, East Coast

Other Recent Posts

New EnergyStar Option for Defining Airflow

In Release 13, we introduced two options for defining airflow through servers based on data from AS…

Read More

How effective is your cooling system?

If one of your air conditioning units (ACU) fails, which part of your data center will be most affe…

Read More