Environmental Monitoring for Hyperscale Data Centers

December 13, 2019 Marc Cram, Server Technology

Environmental monitoring is ubiquitous in most hyperscale data centers. Why? It is due primarily to their reliance on both high ambient operating temperatures and incredible power densities within the IT racks. Their goal is to achieve maximum energy efficiency. To conserve energy, most hyperscalers adjust the set point of their cooling systems to within just a few degrees of the highest possible inlet air temperature that the server and storage equipment can tolerate. This minimizes the amount of time the cooling systems operate, and subsequently the amount of energy expended on cooling.

As such, knowing the air temperature at every IT rack becomes crucial. If the set point is out of sync with the inlet temperatures, equipment can fail. Managing at this level of data granularity presupposes accurate thermal modelling of the data center. It also restricts the cooling system response to the bare minimum, such as opening a louver or floor vent close to the rack that is in danger of experiencing a thermal excursion. As mentioned, the cooling system response must be timely to prevent thermal runaway from occurring.

Maintaining and monitoring a hyperscale data center is no small feat. You are constantly managing the data center’s environment to ensure potential failures are identified as quickly as possible in order to maximize uptime. Numerous methods exist for environmental monitoring in the datacenter. Data center operators have the choice of working with:

1) Wireless sensors offering a one-way communications protocol. Used for flexibility of placement and where the network operations team does not want the sensors on the wired network.

2) Wired sensors that are part of the building management system (BMS). Typically found on the ceiling and underneath raised floors.

3) Passive infrared cameras, like a FLIR-type system that visually show hot spots. These can also work as motion sensors that show personnel on the floor.

4) Sensors that are embedded within the existing compute infrastructure, such as those found inside of the servers of most major OEMs.

Another method for environmental monitoring includes using intelligent data center Power Distribution Units (PDUs). Since they reside within the IT rack, right at the source of most heat generation within the data center, the sensors supported by rack PDUs provide an economical means of providing real-time information. This data is often more accurate than sensors mounted at the ceiling level or within the ductwork of the data center air handling systems.

Intelligent data center PDUs can also support the monitoring network by providing other rack-based data. In addition to temperature data, intelligent PDUs can be outfitted with a variety of sensor types that collect additional information required for the management of the data center. They can provide data on power usage. While feedback on the operation of the cooling system is of utmost importance, so is the physical security of the rack itself.

Adding more sensors and monitoring capability to an already intelligent PDU gives you more visibility into your racks. Armed with knowledge about environmental conditions, your intelligent PDU can help you become more proactive in maintaining not just loads and power, but temperature, humidity and security.

Environmental sensors give you the capability to collect different data points, including:

1) Temperature and humidity

2) Airflow

3) Leak detection

4) Door intrusion and security

5) CCTV

6) Smoke

7) Vibration

After information from data center environmental monitoring sensors is collected by an intelligent PDU, it can be accessed through the rack PDU’s monitoring system. The data can be aggregated and used within the intelligent PDU software or passed along to a DCIM or other building automation system. Rack-based environmental monitoring typically offers the following:

1) Consolidated environmental and power system monitoring

2) Rack-based aggregation of multiple data center environment sensors

3) Improved operational efficiency of the data center cooling system

4) Predictive trending of environmental data

5) Alerting of security or environmental issues on a rack by rack basis

 

Powering Hyperscale at the Edge

Lower latency and bandwidth demands mean hyperscale data centers also need to bring their storage and compute infrastructure closer to end users through edge computing. Remote hardware systems supporting edge usually have firmware and/or software running on them that needs to be fault tolerant, robust enough to be able to run consistently through intermittent communications outages and power availability challenges that result from natural disasters or manmade disruptions.

By definition, edge compute sites are not adjacent to the core data center facility and this lack of proximity means that there is an increased reliance on the ability to monitor power and cooling conditions remotely, as well as the ability to control and remotely reboot single outlets, avoiding the need to dispatch a truck with staff to investigate and correct malfunctioning systems. Edge computing demands more sophisticated monitoring solutions at the rack and PDU level and additional intelligent PDUs features like per-outlet-power-sensing (POPS), enables remote power, temperature, and humidity monitoring which can be used to safely adjust environmental conditions in your cabinets.

Density is an issue when you look at edge computing. The server, storage, and network requirements are high and the amount of space to house them is low. Most deployments

are less than 100 square feet, confined to a 10’ x 10’ space. There are many deployments in which this 100-square foot average is a luxury. The increased computing density leads to higher kW/rack demands, which are quickly followed by a higher plug count and amperage requirements. Innovative rack mount PDUs can overcome the limited physical space that edge deployments allow by providing up to 42 separate C13 outlets in a single, 42U high rack PDU. They can also provide the ultimate in flexibility at the edge with 2-in-1 outlets that can accommodate both C14 and C20 plug types.

In short, an intelligent PDU can provide power monitoring, management, control, and environmental feedback, all in one device. While there are several things to consider when choosing a data center PDU, keep in mind that power is the biggest recurring cost, and the right PDU in the rack can reduce complexity and increase efficiency.

 

Marc Cram is director of new market development for Server Technology, a brand of Legrand (@Legrand). A technology evangelist, he is driven by a passion to deliver a positive power experience for the data center owner/operator. He earned a bachelor’s degree in electrical engineering from Rice University and has more than 30 years of experience in the field of electronics. Follow him on LinkedIn or @ServerTechInc on Twitter.

 

Previous Article
Standardized ARM Modules: The Affordable Way for GUI Designs
Standardized ARM Modules: The Affordable Way for GUI Designs

Do you remember the days when we used mechanical buttons, slide controllers, rotary switches and analogue d...

Next Article
Verification IP: A Vital Component of Chip Design Verification
Verification IP: A Vital Component of Chip Design Verification

To many chip design verification engineers, VIP could easily stand for Very Important Possession and not Ve...