In this week’s Voices of the Industry, Jeff Klaus, General Manager of Intel® Datacenter Solutions, discusses a case study where Intel was able to reduce power cost through analyzing the energy and cooling data across their 60 data centers.
The numbers, however which way you look at them, are writ inexorably large. Worldwide, Cisco predicts about 50 billion devices will be Internet-connected by 2020, while Morgan Stanley, more bullish, projects the number at 75 billion. Cisco also estimates annual global data center IP traffic will reach 10.4 zettabytes in three short years, with data center workloads overall more than doubling and cloud workloads more than tripling over the same period.
All this connectivity has a cost — affecting businesses’ pocketbook and the planet. Worldwide, data centers use about 30 billion watts of electricity. That’s the equivalent output of 30 nuclear power plants, enough for all the households in Italy. Energy costs are the fastest-rising expense for today’s data centers, so it’s no surprise that power consumption is a top concern for managers of data center, enterprise, and cloud environments.
A Complex Challenge
Managing power usage in the data center environment can be complex. The facility team for a particular building may be tasked with measuring and managing power at rack and Power Distribution Unit (PDU) levels, but often has limited visibility into server consumption. Servers typically use 50 percent of a data center’s power, even when those servers are idle. Moreover, there are multiple proprietary power measurement and control protocols supported by various solution providers, making it challenging to have a single solution for power management across all devices in the data center.
Given recent increases in data and energy costs, Intel IT, which operates 60 data center facilities worldwide, sought a standardized data center management solution that is designed to manage complex data center infrastructures, can help reduce costs, and is easy to deploy and use.
To accomplish this, the enterprise decided to deploy a standalone data center management solution that provides accurate, real-time power and thermal monitoring and management for individual servers, groups of servers, racks, and other IT equipment, such as PDUs. In short, a solution that offers benefits for both IT and facility administrators, enabling these groups to collaborate to reduce the facility’s energy footprint. They found it by looking at their own Intel internal teams, and deployed Intel Data Center Manager.
Energy Efficiency, Real-Time Power and Thermal Data
Along with ease of use, simplicity of deployment, and interoperability among diverse server models as well as a variety of products from PDU and rack suppliers, Intel DCM increased energy efficiencies in the enterprise data centers while providing immediate benefits. Real-time power and thermal data for racks, rows, blades, and data center rooms assisted the IT staff to manage data center hotspots and perform power usage planning and forecasting. Receipt of detailed information about server power characteristics helped them set fixed-rack power envelopes and safely increase server count per rack, which improves data center utilization. Aggregating the data for the rack, row, and room with temperature data for the server inlets created a real-time view of power consumed by the servers. Lastly, Intel DCM helped IT staff manage power across various devices from multiple suppliers, eliminating the need for separate equipment-specific tools, and assisted data center managers to make data-driven decisions when maintaining operations during power outages — improving business continuity.
In its initial use of the data center management solution, the Intel IT considered the platform to be focused primarily on better understanding power consumption and the thermal status of servers. However, when it implemented a broader deployment across a greater number of production nodes, the IT staff and facilities operators soon discovered that the solution was also able to help them achieve far more.
Capacity Utilization Based on Power Costs
Real-time data from a large sample of servers enabled more accurate and proactive capacity planning. By virtue of the data center management solution’s historical server power consumption data, IT managers were able to efficiently plan for future server needs. Armed with data that is more accurate than relying on nameplate or de-rated power specifications, Intel IT can now make strategic decisions as to where to invest in new capacity or how to utilize capacity based on power costs.
Secondly, aggregating all the server-inlet temperature data into thermal maps enabled Intel IT to check the effectiveness of its facilities’ cooling solutions and airflow design. For example, it was determined that one data center had a cooling analysis that aligned with the room’s low Power Usage Effectiveness (PUE), and another facility had a cooling analysis indicating that the room was overcooled, which aligned with that data center’s less-efficient PUE.
Next, the data center management solution aggregated control of power devices and helped eliminate the need for expensive smart power strips. Where the facility was already using smart PDUs, their real-time data was integrated into a centralized control panel. It also provided predictive thermal analysis, which helped in several cases to eliminate thermal risks. The solution helped to identify hotspots and thermal inefficiencies that provided data center managers with enough information to change the design of problematic rooms to increase energy efficiency. The cooling analysis function alerted them to hotspots and also helped identify faulty temperature sensors in one of the data centers. Compared to previous manual computational fluid dynamics analysis, this real-time hotspot detection allowed the IT staff to solve thermal issues in the room and increase both thermal and energy efficiency.
Ghost-Busting and Reduced Total Cost of Ownership (TCO)
The console’s automatic identification of idle or rarely used “ghost” servers revealed opportunities for consolidation and virtualization. Besides saving power, this feature helped determine optimal server reuse and reassignment of assets, enabling the Intel IT staff to migrate low-usage devices into virtual servers.
Most significantly, the Intel DCM provided real-time power and thermal data to enable the IT staff make better decisions about load balancing and right-sizing the data center’s environment. It created user-defined alarms that warn of potential circuit overloads before any actual failures or service disruptions occur.
Intel IT and the facilities operations team estimates that this type of data will help them reduce their overall data center energy consumption while improving operational efficiency, thereby reducing TCO.