Intel’s Jeff Klaus on how DCIM tames the data center energy beast

Jerry Bowles Profile picture for user jbowles February 27, 2018
We asked Intel DCIM expert Jeff Klaus how to tame the fast-growing, power-hungry and blockchain-crazed energy demands of the modern enterprise data center

Data center management DCIM businessman iPad © everything possible -
Computers are enormous guzzlers of energy. Based on the most recent Department of Energy (DOE) study, published in 2014, US data centers consumed about 70 billion kilowatt-hours of electricity, representing 2% of the country’s total energy consumption. That's equal to the amount consumed by about 6.4 million average American homes.

The good news in the report is that despite a voracious demand for increased data center capacity between 2010-2014, data center energy consumption grew only slightly — about 4% — which is an enormous improvement from the early 2000s when it was growing by nearly 90% a year.

The energy efficiency improvements are due to a number of factors. These include virtualization, cloud computing, the aggregation of workloads into fewer, but more efficient and powerful servers, and the growth of massive, hyperscale data centers built from the ground up to be energy-stingy. Also important has been the development and adoption of a new generation of data center management tools that visualize data consumption and remotely monitor power and thermal issues in real time.

DCIM to the rescue

That’s where Jeff Klaus, GM of Intel Data Center Manager (DCM) Solution, comes in:

The Data Center Infrastructure Monitoring (DCIM) market is vibrant and evolving quickly. Data center managers need access to granular data that allows real-time power and thermal monitoring as well as health monitoring and utilization. Customers are also interested in quick ROI, and so they are focusing on power management, thermal management, and asset management as areas to unite and provide rock solid process support and analysis.

Klaus explains that Intel’s Data Center Manager Solution is a suite of middleware software that does precisely that. It automates the collection, aggregation, reporting, and logging of a broad range of device status information that is monitored remotely on consoles and data center dashboards. The result can be a significant reduction in operational costs because of insights into the performance of data center hardware and automated responses. Klaus says:

With all of the attention focused on the high-level data center models, data center managers often overlook some of the significant advancements at the hardware level. The latest servers, storage devices, switches, racks, power distribution units, cooling equipment, air handlers, and the myriad of other components are all feeding status information onto the network. Servers, in particular, have also evolved to give remote IT teams many new, fine-grained monitoring and control options.

Cooling and energy costs

Intel’s DCM monitors real-time server inlet temperatures and power consumption data from rack/blade servers, PDUs, and UPSs, and visualizes the data as energy and thermal maps of individual server rooms and data centers.  Klaus says that when combined with control capabilities such as power capping and dynamic server frequency adjustments, the aggregated intelligence helps data center and facilities teams better understand and manage energy costs.

DCM’s companion product, Virtual Gateway, is a set of APIs that let datacenter operators access logs and BIOS data to monitor system health metrics. with a keyboard-video-mouse (KVM) interface.

With an estimated 40% of the total energy used in data centers consumed for cooling IT equipment, even small reductions in energy usage can be significant.  Klaus says the average 300-rack 3MW facility could save 20% in cooling costs from only a four-degree temperature decrease.

Ready for blockchain?

The next surge in demand on data center energy resources may be driven by blockchain applications, such as cryptocurrencies including Bitcoin, which is notoriously energy hungry. An index from cryptocurrency analyst Alex de Vries, aka Digiconomist, estimates that the simple act of ‘mining’ a single Bitcoin consumes as much energy as a year’s average consumption by 26 US households. In a year, therefore, Bitcoin creation consumes roughly the same as the total energy needs of the population of Uzbekistan, a country of 32 million. Klaus notes:

While cryptocurrency companies only make up a relatively small percent of the data center market — for now — what's shocking is the immense amount of energy demand cryptocurrency creates for the industry. Understanding the potential energy demands of the growing cryptocurrency industry must stay top of mind for data center managers as they recognize the need for high performance hardware to solve computational algorithms at a high speed with accuracy and efficiency.

My take

Electrical energy comprises a major portion of the cost of operation of data centers. The insights provided by this kind of software into reducing power and improving thermal management seem to be low hanging fruit that goes directly to the bottom line. On top of that, there is no need to buy new hardware. DCM works with the hardware you already have.

However, a recent study by Intel DCM and Redshift Research found that four in 10 data center managers in 200 facilities surveyed in the US and the UK still rely on MS Excel spreadsheets and Stanley tape measures to initiate expansion or layout changes.

Data Center Infrastructure Management software, such as DCM, provides increased levels of automated control that allow data center managers to receive timely information to manage capacity planning and allocations, as well as cooling efficiency. It can even solve the problem of zombie servers by consolidating servers to reduce energy consumption from 10% to 40%.

A grey colored placeholder image