Can I Get Some Feedback Here?

July 8, 2010

Data Center automation is a worthy and important goal for most organizations. This may be true whether you are running lights out, collocated or just with a lean staff. The more routine operations and processes that can be removed from required human intervention the better. Potential benefits range from improved delivery on SLAs, extended equipment life, reduced labor costs and even enhanced job satisfaction.

Automation systems are simply feedback loops. The components may already exist in your datacenter including sensors, polling and trap management and rule driven software to effect change. Unfortunately the right sensors in the right places in the right quantities are often lacking to complete the loop.

This is part one of a multipart series on automation techniques beginning with heat load management and related sensors.

CRAC / AHU Feedback Control

Below are just a few scenarios that can be a high value target for temperature sensor based automation:

• Scenario One: What if your critical applications are in a collocated facility far away from your IT and facilities staff? How would a cooling failure impact you equipment and more importantly how would you organization react?
• Scenario Two: What if your heat load varies significantly with hours of peak processing?
• Scenario Three: What if an obstruction is blocking or a pulled floor tile is reducing air flow?

Temperature sensors can provide the input necessary to proactively alert and remedy these situations.

Scenario One:

With temperature sensors in the rack and an application that can perform power control we can automatically shed the heat load where temperatures have passed a critical threshold in a zone.

Raritan PX PDUs have this capability when temperature sensors are attached. No additional software or hardware needed. Furthermore we can turn the same systems back up automatically when the temperature is sensed to have lowered to an acceptable level.

Additionally a Raritan PDU can send an SNMP trap alert based on temperature sensor readings to any SNMP trap manager and any SNMP enabled application may perform the same outlet control on the Raritan PDU.

Scenario Two:

Without automation if your heat load varies by utilization you have two choices. Run cooler than needed for much of the processing day wasting energy or run hot and impact the reliability of your equipment. When we introduce automation we can change our cooling output to match the heat load and eliminate the undesirable outcomes above.

Temperature sensors in the rack can provide trigger data to an application capable of altering the CRAC unit output to match the current heat load. A poll of all sensors could be implemented or SNMP alerts could be sent to this controller application based on thresholds such as with the Raritan PDU attached sensors.

Scenario Three:

Increasing compute densities require increasing levels of precision when managing environmentals. Ashrae recommendations are for high, mid and low sensor placement that can facilitate temperature gradient mapping. With a single sensor in a rack an open KVM drawer could go undetected while obstructing airflow to equipment behind it. Similarly missing floor tiles can alter air pressure needed for the cool air delivery. Both air flow and temperature sensors are useful in this scenario.

Raritan PDU connected sensors have associated X, Y & Z coordinates to facilitate a proactive approach for immediate identification and location of these types of situations.

The next feedback loop automation we will look at next will focus on applications.

Jonathan Davis