This is the second in a series of blog posts providing tips and tricks for answering common data center questions with the use of SPM. Last month, I gave some guidance on answering, “How do I identify the best places to install new IT equipment?” Once you understand the ways SPM can help you with that question, the next question tends to naturally follow.
Question: How do I predict when I will run out of power?
The importance of the answer to this question is obvious. It is a primary input for growth and capacity planning in the near-term and long-term future of any organization with a data center.
- Understanding today’s power usage is the first step. SPM offers several levels at which to monitor the power usage from the granular outlet-level measurements of the POPS CDUs to the overall building-level Locations.
- Cabinets are a useful monitoring point for the obvious reason that the IT equipment being installed must fit within the capacity of the circuits brought into the cabinet. Setting up cabinets in SPM also provides a redundancy check for A/B power feeds in addition to comparison of power draw to capacity.
- Circuits are a means to aggregate current and power levels of multiple cabinet CDUs into hierarchical levels that are directly comparable to inputs of upstream power devices such as RPP, PDU, and UPS. Setting up circuits in SPM provides Smart access to dumb upstream power devices.
- The optional key-activated Custom Devices feature of SPM provides the ability to define SNMP capable upstream power devices. By entering the OIDs of input and output current, voltage, and power, the SPM can monitor the usage as reported by Smart RPP, PDU, and UPS devices.
- Locations are the top level of monitoring when trying to understand today’s usage in the organization. By completely setting up a meaningful hierarchy of CDU to Cabinet to Location(s) with Circuits and Custom Devices added, the data center is defined so that growth and capacity planning can move forward.
- While setting capacities for the items in “1” above, alert thresholds for total power of those Cabinets, Circuits, and Locations can be set as well. See the previous post for more tips.
- Alerts on Current measurements are needed to prevent tripping of circuit breakers or fuses. The safety rated levels for current on the CDU and Circuit phases are critical for this purpose on a moment-to-moment basis.
- Alerts on Power measurements are needed to understand proximity to general capacity limits when comparing equipment power usage and cooling requirements. This is typically useful at the Cabinet and Location levels above all else.
- Ultimately, the data center and facilities teams want to predict when power at each level will reach its limits. Knowing when growth will trigger the need for more resources is important for any rapidly growing organization.
- Predicting the future is a challenge, no doubt. The first thing to understand is past experience. By setting up Cabinets, Circuits, and Locations, and then waiting and watching for a sufficient period of time, one can then identify growing power usage at any particular level in the data center. The period of time to monitor is dependent upon the individual situation. There will typically be daily, weekly, and monthly cycles of power usage; however, the usage over several months will typically show the overall general trend.
- The predictive trends provided by SPM for power give two linear fits to differing time frames. For example, you may look for increasing usage based on 3 month and 12 month historical data. Additionally, like the standard trends, the predictive trends state the minimum, maximum, and average measurements over the given period of time.
- By setting up alert levels in #2 above, predictive trending analysis can extend to alerts on future conditions. On a per-device (Cabinet, Circuit, Location, etc.) basis, this alert can be activated for notification of a potential breech of the threshold within a specified period of time in the future.
Understanding growth, past, present, and future, is an important aspect of running the data center. For more information on Growth Planning, see the Application Note on the subject.