Real-time streaming data stored with timestamps allows users to understand the baseline capacity loading, temperature and power consumption profiles over time says Dr. Prajesh Bhattacharya of OSIsoft Electrical Data Often all of the power used by the critical IT equipment fl ows through the UPS. Each has specifi c effi ciency vs load characteristics and typically, the greater the operating load, the higher the effi ciency. It is important to keep track of the load and the effi ciency to check if it is following the vendor supplied characteristics. Operate one UPS at a higher load rather than operating multiple UPSs at lighter loads. IT service load is dynamic in nature and consequently the IT power load is also dynamic to some extent. Therefore, it might not be benefi cial to control the operation of UPSs based on just the last snapshot of the real-time data, because that might lead to switching load from one UPS to the other at a frequent rate. Instead, scheduling the IT power load to bypass a particular UPS for a certain period of time would be a more robust approach to run UPSs at higher loads. This will be possible only if one has the history of the UPS power draw.
Mechanical Data
In data center critical facilities, chillers represent a substantial capital investment and are a major contributor to operating costs. Every chiller has a vendor supplied design effi ciency (kW/ton) fi gure. If the historic data shows that this value is consistently higher than the design value, then one can conclude that the chiller is under-performing. Refrigerant Approach Temperature (RAT) is the fi rst thing to be checked for heat exchanger (either evaporator or condenser) inefficiency. When the design value of the RAT is exceeded on a regular basis (which can be known only from historic data), there is a problem with the heat exchange process. For a high pressure chiller system, if the refrigerant charge level decreases over time, there is a leakage in the system. One should also check for leakage if the evaporator pressure is less than the design value. For a low pressure chiller, if the measured condenser pressure is consistently greater than the design value, then one should check for non-condensables and fouling inside the condenser tubes. The purge-unit runtime is another indication of moisture inside the refrigerant system. If it runs too frequently (check the total runtime over the last week, last month, last 3 months, last 6 months for example), then one should look for moisture leakage into the system.
Most data centers have redundant chillers that operate all the time at light loads. It is always advisable to operate fewer chillers at higher loads and to turn on the redundant ones only when it is required.
If the amount of dissolved chemicals or particles in the cooling water is consistently greater than the recommended high limit, check the water treatment system as poor water quality can foul the heat exchanger. A consistent increase in the pressure drop across the cooling water fi lter indicates greater particle density of the supply water. The filter should be changed or cleaned if that pressure drop exceeds the recommended high limit. Greater pressure drop across the oil fi lter indicates change in the oil characteristics, which can give rise to various problems. Monitor the pressure drop across the oil fi lter on a continuous basis.
IT Data Benefits of monitoring IT systems include improvements in application availability, network performance, SLA compliance and capacity planning.
With regard to IT, who needs real-time data, historical data and why? - Capacity planners who struggle with retrieving enough relative information about application performance to make informed decisions about future capacities. DEG. Security and system analysts that are faced with a myriad of fragmented node and server log files. - Application analysts need to track and review application utilization statistics determining design fl aws and excessive or unauthorised utilization. - The CIO/CTO who must understand the overall effi ciency and throughput of the IT infrastructure through a highly available and visible operations data infrastructure.
Holistic Critical Facilities Management IT uptime, power usage, cooling, alignment between operational cost incurred, charge-back to the customer and capacity management at key measurements. Some of these areas are typically parts of a Service Level Agreement (SLA). Someone operating several data centers might be interested in all of this information for all the data centers.
The left pane of figure 1 shows a drill down from continent via country, city, data center, colo to zone and the right pane shows real time and historical data.
With real-time data management infrastructure, it becomes possible to correlate IT with electrical and mechanical information in real time and historically.
Technology is available today to implement an enterprise operations data infrastructure. Best practices in process manufacturing industries, such as electric utilities, are well known and accepted. Looking at the data center as an "IT Plant', real-time infrastructures will have a similar, significant impact to help users optimize operations.
