Data Center
Energy Analysis
Architecture
Contents
1. Executive Summary
2. The High Price of Low Efficiency 3. Sentilla Puts Energy in IT Management 4. The Inputs: Cost, Work, Context 5. Automated Analysis
6. The Output: Added Value for IT Management
7. Summary 8. About Sentilla
Understanding energy usage throughout a data center is critical to good IT management practices and cost reduction. Yet, energy management software to evaluate and monitor data center performance, capacity, and behavior is not widely used. The Sentilla Data Center Performance Management (DCPM) platform supports CIOs, IT managers, and data center strategists by presenting key performance indicators (KPIs) and operational insight.
As an integral part of an IT management portfolio, the Sentilla DCPM platform collects real-time energy data that characterizes the state of the data center, and high-value analysis that suggests improvements. As such, it is an invaluable tool to assess the initial efficiency of a data center, and to guide and track its evolution toward lower operational cost.
The Sentilla Analysis Server, which forms the corner stone of Sentilla DCPM, delivers actionable intelligence about the data center’s energy consumption. From real-time alerts that prevent outages, to trend tracking and capacity planning, the Sentilla DCPM platform provides essential insight to IT managers that help them reach their goals and document their achievements.
The High Price of Low
Efficiency
It is estimated that, by the year 2020, the carbon footprint of the world’s data centers will surpass that of the airline industry1 due to their electricity
consumption. The cost of running a data center promises to rise as fast as the threat of climate change. In 2010, energy expense in data centers is projected to exceed that spent to buy new servers2. Over the lifetime of a data center, the cost to run IT equipment can easily exceed the initial price tag by a factor of ten.
Green policies aim to curb energy consumption and CO2 emissions through the implementation of energy efficient techniques. These techniques
include, among others, proper thermal isolation for optimal cooling, turning off servers when not in use, and virtualization.
However, data centers can be vast and complex. This makes it difficult to identify where cost is running high, if the data center has capacity left for expansion, or what servers are good candidates for consolidation. All these issues are ruled by their energy aspect: where it is wasted, if there is enough, and how to get more work done with the same amount of energy.
Therefore, IT managers need to view energy as a limited resource in their facilities, like storage and computing capacity. Tools are required to automate the monitoring and reporting of energy consumption and capacity.
Sentilla Puts Energy in IT
Management
Information about the performance, capacity and cost of IT is already available to data center managers. In fact, there is sometimes so much information that it has become difficult to make sense of it, which has led to the rise of enterprise management software suites. But no such IT tool exists for managing energy resources. Energy data, often collected from different sources in the data center, is typically incomplete and noisy. Worse, there is no standard place to view and analyze key energy indicators.
The Sentilla DCPM platform gathers information from different sources within the facility (multi-vendor infrastructure, equipment, system software, blade servers, etc.) and consolidates data into one place. With this information, the Sentilla DCPM platform digests and presents actionable intelligence about the data center. At the core is the Sentilla Analysis Server, shown in Figure 1. It examines the data center’s incurred costs (energy use, equipment, and cooling profile) and work performed (applications, transactions, and CPU load). The Sentilla Analysis
1 - Mc Kinsey & Co., “Revolutionizing Data Center Efficiency”, Report, April 2008.
Server continually processes cost and work data in conjunction with business rules and context specific to each data center (function, owner, capacity, hierarchy, and location). The result of this sophisticated real-time analysis is the discovery of areas of waste, bottlenecks, abnormal behavior, and potential performance and cost improvements. This makes the Sentilla DCPM platform an essential tool for IT managers who already leverage software that analyzes performance, application usage, and trends in their network and applications. The Sentilla DCPM platform adds automated management of energy performance, capacity and cost to IT manager portfolios. For instance, while IT managers may use Wily technology for application performance, or Tivoli for network and server management, Sentilla provides the IT tool for energy performance and management.
Consisting of a robust data collection platform, analysis server, and business rules for data center management, the Sentilla DCPM platform takes energy data and turns it into actionable IT information. The following describes how content is analyzed and presented by the Sentilla Analysis Server, the core component of the Sentilla DCPM platform.
The Inputs: Cost, Work,
Context
The added value of the Sentilla DCPM platform is its capacity to automatically understand work and cost in the context of a data center. To do so, the first task is to find out what is being used in the data center and to allow proper profiling of the data center’s energy usage.
Operational Cost
To manage energy costs, IT managers need information about equipment in the data center, which must be gathered from multi-vendor infrastructure. Generally speaking, the collected inputs are “sensed” through various means: dedicated power meters, temperature and humidity probes, already running software agents, etc. This real-time data directly translates to a cost: server operating expense, cooling increases due to temperature changes, power distribution loss, or changes in humidification.
Electricity measurement tools or built-in hardware on server equipment automatically characterize the energy consumption of assets. Wherever data is incomplete, the Sentilla DCPM platform can infer
Utilization Analysis Real -time Analysis
Continuous Analysis Report Engine Asset Profiles Meters Meta data
Sentilla Analysis Server
Real-time Data
Immediate Alerts
Historical Data &
Trends
RecommendationsReports
Asset & Configuration Databases Servers, Storage & Infrastructure Enterprise Management Tools Time Seriesunmeasured values by understanding the hierarchy of the data center. For instance, in the topology of Figure 2, the power distribution loss at the UPS can be deduced from the available measurements. These elements assess the cost (both monetary and environmental) of each operation, therefore providing insight on the macroscopic business view of each facility.
Work
Next, look at the amount of service and work being done in the data center and correlate this activity to the energy and cost data that was collected. Enterprise management software typically records the IT load experienced in the data center. It represents the amount of work performed, that is, computing resource usage, services, transactions, and applications—the business functions of the data center.
Taking IT work and putting it into context can answer important management questions: where to install new servers to balance power capacity? Does the distribution of work across servers create power hotspots? Which equipment should be consolidated while maintaining SLAs?
Context
To make sense of the cost and work information, it must be put in context. Each data center supports the business activities of the company it supports, creating a unique set of priorities and uses. This impacts applications, layout and composition of the equipment, and performance requirements. This configuration and metadata provide context for the raw resource usage collected from the infrastructure and enterprise management software.
For instance, assets placed in the same logical group can be compared to one another, and outliers may be identified, suggesting that the equipment is either poorly configured, or underutilized. Asset metadata provides information to put energy consumption in perspective: a consumption of 1,000W may be alarming for medium-range servers, but is expected of high-end blade servers.
Although collected data contains a wealth of information, it is often incomplete, noisy, and hardly manageable. The Sentilla Analysis Server automatically sorts through the data, correlates it with contextual metadata, and extracts actionable content.
Automated Analysis
All of these inputs are directed to the Sentilla DCPM platform to be stored, filtered, and analyzed. Sentilla DCPM then channels real-time data through a message queue and off to web services and a web dashboard to be displayed for the end user. However, this data is also examined by the Sentilla Analysis Server, whose role is to sort out past and present data in succinct and meaningful summaries and metrics.
Figure 2 - Aggregate data can help determine power for non-measured assets. For instance, the power entering the UPS can be computed as the nominal switch gear power (P0) minus cooling power (P1).
Real-Time Evaluation
Analyzing the stream of energy data in real-time provides immediate feedback on critical operations of the data center. With thousands of data points arriving every minute, real-time evaluation extracts key urgent alerts and delivers them via multiple means (SNMP, email, SMS, etc). These alerts can notify operators when a phase is overloaded, a power supply fails, or energy consumption is near capacity.
Since not everything in a data center may be directly assessed, unmeasured points can be estimated using partial data in conjunction with context, power hierarchy, and management information. With measured and extrapolated data, the analysis server provides a real-time view of the complete data center, including where energy capacity is used at any point in time.
Additional information can also be derived from the power values and energy cost. The Sentilla Analysis Server computes current and past energy consumption, the associated cost by time of day, and the corresponding carbon footprint in kg of CO2. These values can form the basis of a per-Watt cost study of all operations performed in the data center, providing insights for business optimization.
Continuous Analysis
Beyond real-time heuristics, continually analyzing energy information in the context of the work that the data center is performing identifies abnormalities and trends, assists in capacity planning, and avoids risk. Through the use of metadata, such as server profiles, the outliers within a logical group can be identified. Every asset has a power “signature”, a unique profile of consumption validated under normal operating conditions. With this correlated feedback, IT managers can single out underutilized assets, misbehaving equipment, or favorable candidates for consolidation.
Furthermore, by determining asset profiles, the Sentilla DCPM platform detects equipment that deviates from its usual operational behavior using statistical analysis. Over time, equipment exhibits
predictable trends, which can be difficult to track manually. As trends may point to performance degradation, this knowledge allows IT managers to become proactive and take preventive measures such as replacing faulty hardware or shifting applications away from assets nearing capacity. It also identifies abnormalities and outliers that come from degraded performance, inefficient configuration, or preliminary signs of upcoming failures.
With a library of potential improvements or policy changes, the data center’s operations can be mapped to relevant recommendations. To evaluate the priority of each proposed change, their impact on cost, ROI, and disruption can be calculated. Ultimately, these changes factor into the overall data center strategy and execution plan.
The Output: Added Value
for IT Management
Continuous analysis gives IT managers a new set of tools to reduce or avoid downtime and increase performance. The resulting outputs can be accessed via any number of means, whether web dashboard, SNMP, web services, or email. We describe their importance in the following.
Real-Time Data
The Sentilla DCPM platform dashboard displays real-time data for quick management and review. It shows an overview of the energy consumption per resource and facility, with its equivalent in energy costs and carbon output. Key performance indicators, such as utilization, cost per petacycle and PUE allow IT managers to gain immediate knowledge of the data center’s conditions, performance and efficiency.
Because data centers can be very large, the Sentilla DCPM platform nests information in a hierarchical manner to ease the viewing experience, allowing IT managers to map energy performance to function or physical configuration.
Archives
Besides displaying real-time data, Sentilla DCPM maintains historical data. This helps establish the evolution of the data center over time. Archives can also interact with other tools, for instance to be used for billing purposes or internal chargebacks.
Alerts
A core feature of Sentilla DCPM is the display of findings from the analysis server. Today, most energy data is processed manually in Excel spreadsheets and text files. That human-intensive process can take days, weeks, or even months. Instead, Sentilla DCPM immediately informs IT managers of potential problems in the data center. IT managers can gauge right away the severity and the extent of a problem. By reacting quickly, they can reduce downtime, or avoid it completely.
Alerts are thrown in the case of critical operation events, prolonged idling, low electrical capacity, or unfavorable budgetary developments. The Sentilla DCPM platform watches over all energy aspects of a data center. These alerts may be delivered via a variety of means – SNMP traps, email, and web dashboard notifications to name a few.
Recommendations
Recommendations suggest improvements for IT managers, allowing them to go farther to lower cost, free up capacity and improve performance. Sentilla DCPM provides recommendations to improve energy efficiency based on the asset profiles and trends developed by the analysis server. It also examines the results of statistical analyses to bring attention to outliers and abnormalities that lead to inefficiencies or risk of outage.
Continual analysis uncovers idle periods and determines if an asset is unused, partially used (depending on the time of day), or marginally used. This information is useful for identifying candidates for consolidation.
Asset profiling also enables comparison of similar equipment, which offers IT managers a chance to review the workload balance in a group of assets. Recommendations include budgetary information combined with strategies for improved performance, such as server and storage consolidation, storage optimization, equipment replacement, and facility improvements.
Reports
With the use of industry standard reporting packages, data, trends, and analyzed content document the progress made toward lowering costs, reclaiming capacity, and achieving higher energy efficiency. Further investigation into data center performance is possible by slicing and viewing the data in custom formats. Annotations can be viewed on top of data and trends, providing context for changes in cost and execution.
Reports break down data center cost by largest consumer, an important feature for accurate billing of customers and internal chargebacks, which motivate every actor of the data center to move toward higher efficiency.
The evolution of key performance indicators, including the PUE, demonstrates the changes in efficiency over time and can be correlated with external effects (such as outside temperature). Finally, IT operators track previous actions taken in their data center as the Sentilla DCPM platform prints a history of alerts, recommendations, and the aggregate log of the responses to these problems and suggestions. Similarly, IT managers can track and manage the workflow of actions using their existing ticket tracking system.
Summary
The Sentilla DCPM platform brings all aspects of energy management to the IT managers’ tool portfolio. It groups and analyzes inputs from different sources in the data center to provide real-time insight and actionable intelligence. The Sentilla Analysis Server helps IT managers improve performance and efficiency, reduce downtime, and plan for expansions or consolidation. The aggregated information provided by the analysis server helps shape coherent energy efficient policies into an extensive plan. Once implemented, the success of this plan can be measured and documented clearly. IT staff, data center strategists, and CIOs can leverage energy management as a key component of their IT software portfolio to deliver exceptional IT performance at minimal cost and without disruption.
About Sentilla
IT data center professionals rely on the Sentilla Data Center Performance Management (DCPM) Platform to achieve asset-level utilization intelligence across all their data center and collocation facilities. The Sentilla DCPM platform combines sophisticated performance analysis with on-going “what-if” scenarios designed to optimize capacity planning, modernization, virtualization, and power consumption initiatives. With no meters or agents to install, Sentilla’s award-winning technology implements quickly and delivers a proven annual ROI of greater than 25 percent. Sentilla’s unified view of data center operations ensures always-on availability, cost optimizatialways-on, and calways-ontinuous performance improvement within existing data center infrastructures and budgets. Focusing on achieving the high performance data center, IT/data center professionals are best positioned to align their initiatives with critical business goals.