Data Center Measurement - Building An ROI Baseline
Establishing a baseline measurement of your data center is an important first step in developing a meaningful ROI for any data center project. The cost of energy has recently exceeded the cost of IT equipment; understanding your energy footprint in terms of power consumption and cooling demand has become an important management tool.
to view the recorded webinar presentation, please visit http://www.42u.com/measurement-techniques-webinar.htm
Provides actual data that can be used in data center design and management tools. Create a roadmap Ensuring you’re not exposed to SLA failures??? Ensuring compliance with your ITIL compliance in areas such as supplying appropriate capacities, as opposed to sufficient capacities. More deeply, business analysis best practices, as defined in the Business Analysis Body of Knowledge (BABOK), define the techniques and ITIL v3 where the focus is on the life cycle approach. Ineffective business analysis costs U.S. businesses at least $30B every year. Improving the precision of the Total Cost of Ownership calculation to be used for informed planning. As-configured is not typically the actual experience. Blockages and inefficiencies degrade the actual performance. Identifying stranded capacities that are opportunities for efficiency improvement programs. Gartner recently reported findings (RN-G00153396) that included strong recommendations that IT improve their efficiency by first measuring their energy efficiency losses. Gartner describes operational maturity as moving from chaos through reactivity, proactive services to generation of business value. One of the first steps to take to move closer to and to maintain IT business value is to measure your data center against a meaningful benchmark, such as power usage effectiveness (PUE). Stratavia recently produced an excellent whitepaper describing this process. A measurement initiative can be the basis for a more broad greening initiative that generates considerable savings, frees budget for other efforts and contributes to corporate environmental initiatives. (need green corporate story). The Department of Energy (DOE) will soon release DC Pro, a tool that can be used to calculate data center infrastructure efficiency (DCiE) as well as provide recommendations for improvement.
The A and B projections are for improvements in server technology efficiency.
Power Usage Effectiveness is an established benchmark that can be used to compare your data center efficiency to the national average as well as others in your industry. Power use efficiency is an alternative use of the acronym. EPA uses the term EER, energy efficiency ratio, in their recent study.
The largest opportunities for efficiency gains from upgrading to best practices are in the area of chilled water and fans. Air and Water side economizers are two best practice approaches that will reduce chiller expenses.
You need to target the meaningful level of precision that is cost effective. You can do some gross calculations using room-level data, but that will not yield sufficiently actionable results. Conversely, measuring down to the device level is typically not cost effective; very nice information to have, but in most cases it’s not required. The most effective level is at the rack. Depending on the volatility of your data center you can decide to have a point-in-time assessment, establish an ongoing monitoring environment or invest in real-time monitoring technology. It has been our experience that data centers are rarely static environments, but some purpose-built data centers do have sufficiently low volatility that a point in time measurement approach is the most cost effective approach. The cost and availability of real-time monitoring technology has improved dramatically and is now to the point that it is a very cost effective alternative. It is important to note that there can be value in having device-level data. It can identify unusual airflow requirements, power consumption characteristics of a device, as well as more accurate assessments of the effect of rack configuration changes. In general, this level of detail does require a mature IT environment where change history and inventory management data is maintained. Another factor to consider when determining your measurement instrumentation plan is your need for data to use in a modeling tool that can predict the results of changes before you make the changes. This brings us to an important topic, visualization…
systems & network administrators have tools for visualization useful for debugging, benchmarking, capacity planning, forensics data center managers have comparatively poor visualization tools
… and now for an illustration of the power of visualization.
Computational Fluid Dynamics
Several decades old ~5000 sq ft 18” raised floors ~.5MW total power consumption Combination of clusters and business systems clusters account for almost all new demand Unusual cooling system 7 down-shot CRACs for under-floor supply supplementary overhead supply (chiller coil + fans) Starting summer of 2007, IT began a data center efficiency effort In collaboration with researchers in EETD, we perform a series of engineering studies documenting electrical, mechanical systems measuring loads & capacities; performing CFD calculating efficiency identifying opportunities for improvement deploying wireless sensor system (more later) The data center was overcooled but this is the norm Abundant low-hanging fruit simultaneous humidification & dehumidification blanking panels missing excessive number of perforated floor tiles air-flow pathologies (mixing, short-circuiting) The data center was relatively efficient! none of the clusters backed by UPS
We installed a wireless sensor net from SynapSense 200 monitoring points temperature, humidity, under-floor pressure, current For the first time, we have a detailed understanding of environmental conditions in the data center real-time and historical graphs underlying database
effect of adding one 12” blanking panel to the middle of a rack conventional wisdom is correct: blanking panes are essential other data center clichés also borne out eliminate leaks in floor manage floor tile permeability
Feedback on floor-tile tuning in the course of one day 24 perforated floor tiles removed 6 floor tiles converted from high- to low-flow 4 floor tiles converted from low- to high-flow
Feedback on floor-tile tuning: with instrumentation, we can observe results in real time when airflow is restricted, under-floor pressure increases rack-top temperatures decrease without monitoring and visualization, this process is guesswork how many tiles to remove?
Visualizing impact of site modifications Graphing impact of major maintenance (redirection of overhead cold air supply underfloor) on air pressure in some areas, under-floor pressure increased by almost 50% impact varies according to distance from new air supply
Visualizing impact of maintenance: each CRAC turned off in turn, for service hot spots monitored during maintenance two-fold benefit real-time feedback enhanced knowledge of data center redundancy characteristics provisionally turn off one or more CRAC units?
Get electrical, mechanical studies Find engineering firms with data center experience Consider instrumenting your data center to enhance visibility Eat your spinach (blanking panels, leaks) Take advantage of outstanding, free PG&E Pacific Energy Center classes (state-wide) Keep an eye on emerging technologies (flywheel UPS, rack-level cooling)