The collective voice of IT and facilities teams provide invaluable insights about data center technology, both positive and negative. As general manager of Intel’s data center solutions business, I have had the opportunity to meet and discuss issues with hundreds of these data center professionals. I’m often asked about the common themes and topics raised and what predictions I can make based on empirical feedback.
Over the last decade, telco data center managers, for example, have consistently pushed technology to the limits to keep up with exploding traffic, service growth, and market dynamics. These early adopters of data center solutions have always been excellent solution collaborators, bringing hands-on knowledge of real-world use cases to discussions about changing designs, approaches, and best practices in the data center.
The top challenges consistently described by telco data center managers today include:
- Maintaining oversight of assets in the data center, and understanding how their use varies by application, tasks, and enduser profiles
- Understanding the potential opportunity for refreshing hardware devices, and quantifying how refreshes will impact data center OpEx
- How to gather and aggregate more data center telemetry without increasing hardware complexity (i.e., without introducing more devices/sensors)
In any highly dynamic industry, each of these challenges can significantly impact business results. I continually hear from telco IT and site managers who describe complex server labs with more hardware than they feel they can effectively manage. The fact that literally all of these professionals are managing infrastructure that includes many older assets makes it even harder to determine what should be replaced, refreshed, or left as is.
So how are the world’s most successful telcos addressing these challenges and what can other industries learn from their example?
The ability to tap into power and temperature telemetry data is literally turning these areas of frustration into competitive strengths for fast-growing telcos. The ability to view real-time energy and thermal patterns and to analyze those patterns for varying workloads makes it possible to immediately identify the under-utilized assets as well as those nodes that are running at or over the recommended limits. IT can measure the exact power required for unique use cases, and therefore accurately plan capacity and expansion.
The telemetry data exists already, and there are many solutions available that make it easy to gather and exploit this information without adding any additional hardware. Some examples from Intel’s telco customers include:
- Improving thermal profiles. Türk Telekomleveraged thermal and energy data to safely increase server room ambient temperature by 1.5 degrees. The savings in the pilot deployment added up to $40K per year, and lowered carbon footprint by the equivalent of taking 40 vehicles off the road.
- Increasing rack density. The visibility gained at Chunghwa Telecomallowed the telco to increase rack density by up to 83% without raising the power envelope.
- Reducing power consumption. At Telecom Italia, introducing middleware to aggregate power telemetry data showed them that they could reduce overall energy consumption by up to 20% in the data centers. The increased visibility also pointed out the most inefficient servers as candidates for hardware refresh.
These are just a few examples of how telcos are using energy and thermal management to drive better decisions about overall data center layout and workload distribution. Other highly dynamic and fast-growing data centers are also evolving data center practices for more agile resource management based on real-time conditions.