Dynamic Energy Optimization and DCIM

To justify the high cost of investment in a DCIM solution, most enterprise DCIM software promise energy savings without providing an actual means of doing so.

It appears that monitoring and providing visibility to energy performance metrics such as PUE is somehow automatically equated to energy savings. However, to accomplish energy savings a DCIM system must also be able to physically influence the environment in some manner.

This means that a DCIM solution must include a control system to dynamically control the infrastructure to adapt to the ever changing power and cooling requirements.

Furthermore, the cooling infrastructure and heat dissipation ecosystem within a data center is typically complex, involving many actors and multiple parameters that influence the process.

The control system must understand the dynamics of the physical process involved in order to model the real cost of cooling and optimize energy consumption. This article discusses an advanced control solution that can be incorporated into any DCIM solution to effect energy savings.

Where are My Promised Energy Savings?

A distinct perception in the data center industry today is that DCIM (Data Center Infrastructure Management) is an oversold term. A valid question to ask is why there is this almost negative perception in just a few short years, even before the DCIM market place has a chance to mature?

The simple answer in this blogger's opinion is that so many of these DCIM products have been long on promises and short on delivery.

Since the shiny new "DCIM" industry came into existence about 5 years ago, reputable analysts and research firms have assessed the market value in the hundreds of millions of dollars, with many predicting a value north of $2B by next year 2016.

Naturally several vendors have jumped into the fray to get a piece of the action, almost without exception with existing products rebranded as "DCIM" software. Consequently we now have the somewhat comical situation where DCIM software range from simple Excel spreadsheets with specific functionality, to traditional environmental and power monitoring (BMS) systems, to nebulous "enterprise" systems with obtusely defined functionality.

To make sense of the wide ranging claims prevalent across the new product packaging (let's just stop pretending that this is a new industry OK?), some organizations like 451 research and Uptime have tried to come up with a core set of features that a DCIM software should have. You'd have to consult these references for more details, but according to a 2013 Uptime Institute survey, the feature set include:

1. Real-time environmental monitoring

2. Real-time power monitoring

3. Trending and analysis of historical operational data

4. Capacity planning for power, cooling, and space

5. Tracking IT asset inventory

6. Identifying under-utilized servers and IT devices

7. Real-time cooling optimization

8. Modeling and simulation for airflow

9. Modeling and simulation for equipment changes or operational changes

10. Cost modeling

These are a lot of features. Implicit in this list is that the software has to be vendor-agnostic to have any chance of practical use in a typical data center environment with diverse population of equipment from multiple vendors. Clearly any one software package that delivers on all or even half of these features is bound to be very expensive by definition. Which is where the long-on-promises and short-on-delivery reputation and skepticism surrounding DCIM come in. To justify the expense charged for DCIM software – especially the “enterprise” behemoths – all DCIM vendors promise a monetary return on investment whether they can justify such promises or not.

Looking at the list of features above, the only ones that have provable ROI are the ones that have to do with energy savings. In a production (read operational) environment, this turns out to be a very short list:

1. Real-time cooling optimization

2. identifying under-utilized servers

This of course is not to say that all the other features are not equally as important, but these are the only two that can objectively be monetized. Environmental and power monitoring are crucial in a data center environment, as is asset tracking especially in large facilities. Capacity planning is also very important especially for colos and higher tier facilities. Unfortunately though, these features do not yield the hard numbers to include in an ROI or TOC argument. Which is why many vendors resort to claiming “energy savings” without a demonstrated manner that it can be achieved.

You Cannot Optimize What You Cannot Control

Quite often you hear the refrain “You cannot optimize what you cannot measure” as justification for a monitoring system. That is a valid argument of course, but equally often the obvious other side of that coin never gets mentioned: “OK you can measure everything, now what?” If a DCIM system is promising energy savings simply because it can measure and track power and temperature, that’s a slight of hand at best. Implicit in this claim is that the user would then take a look at the trend and decide – on a day-to-day basis no less – what to do to cut down on energy consumption. Bunch of baloney right? My practical experience with most users is that they just want a simple dashboard showing the status, preferably colored Green if all is clear and Red if there is an abnormality. They only get to take action if something on the dashboard turns red, as simple as that; which is understandable as their typical day is split among a host of responsibilities. A simple monitoring system can accomplish this – it does not have to be complicated and certainly none of this has to do with energy savings.

Here is a new refrain when it comes to energy savings: “You cannot optimize what you cannot control”. To accomplish energy savings, a DCIM system must also be able to physically influence the environment in some manner. This boils down to being able to interact and control elements of the power and cooling infrastructure in some automatic and proactive manner in order to bring about certain desired outcomes. The human operator must not be required in this interaction, except perhaps occasionally and in some pre-emptive fashion. Practically speaking, the most effective way to save energy in an ongoing basis is to set up an automatic control system to do it, and forget about it.

ActiveCFD: Intelligent Dynamic Cooling Optimization Out-of-the-Box

A traditional feedback control loop, which many BMS systems already have, are programmed to respond to abnormalities by supervisory control of the power and cooling infrastructure. In supervisory control an operating point is chosen, and the controller simply tries to maintain that condition by adjusting set points in the cooling infrastructure prime movers.

To accomplish energy savings, especially in a complex critical infrastructure like a data center, a control system must take the traditional supervisory control a step further by incorporating a higher level of intelligence regarding the physical phenomena it is trying to control. Only then would it be able to define an effective set of metrics, and then optimize the energy performance around those metrics.

In any physical system, especially one with multiple degrees of freedom, optimization can usually be accomplished in multiple ways. In a data center environment, it is this author’s opinion that no level of intelligence beats a heightened knowledge of airflow management within the data center; an optimizing controller with this intelligence is the nirvana as far as an effective implementation goes. ActiveCFD™ is such a system. By incorporating high-fidelity computational fluid dynamics analysis within the traditional supervisory control loop, ActiveCFD spots a repertoire of knowledge and metrics that otherwise would not have been possible. Some of these include:

• Smart Zones

• Predictive Monitoring

• Predictive Control

• Automatic Flow Balancing (of containment systems)

• Touch-Button What-If Analysis for Effective Planning and Diagnostics

ActiveCFD utilizes these features to ensure that the controlled environment is always maintained at an optimum operating state – no more and no less of the required cooling is delivered – while at the same time responding effectively to abnormal events. ActiveCFD is deployed with sEnergy EMS DCIM software normally, but it is designed to fit into any existing DCIM or BMS solution with minimal disruption. With this tool your preferred DCIM solution can finally deliver on the promised energy savings depending on the existing hardware infrastructure. For more information and to see if this tool is right for your situation, download the whitepaper