Data Center 2025, is an industry-wide initiative spearheaded by Emerson Network Power to create a vision — or multiple visions — of the future of the data center included a discussion around improvements in data center management. To categorize progress on the management front so that results could be quantified, the major stages in the evolution of data center management were defined as full visibility across all systems and layers, self-optimizing, and self-healing.
A thought provoking statistic resulted: 43% of participants said they expect the industry to move all the way to self-healing, providing a level of automation and control that will enable an unmanned data center in 2025. According to one participant, “Data centers will be highly automated. All devices will have automatic identification tags and data centers will be operated without human resources or with minimum human involvement.”
Participants in Data Center 2025 almost all agreed that progress will be made in data center management and control, but there was no clear consensus on how far it will advance. Here, experts from Emerson Network Power and DLB Associates champion each end of the spectrum.
A Revolution in Data Center Management is on the Horizon
by Steve Hassell, president, Data Center Solutions, Emerson Network Power
There is increasing pressure on the data center to achieve new levels of efficiency without compromising availability. Management platforms are going to be key to accomplishing that, and I expect self-healing and self-optimizing data centers to be viable by 2025.
If you define self-healing as the ability to adapt to failures without impacting application availability, self-healing capabilities already exist.
Thanks to virtualization, many organizations have the capability — and the capacity — to shift loads within the data center without impacting availability. What they lack is the visibility across systems to understand the impact of those changes and adapt support systems based on them.
That’s why the “single pane of glass” that DCIM promises is so critical to the future. The primary technical challenge is that the data center is a Tower of Babel — every system is communicating in a different language.
Until recently, there hasn’t been a way to translate the data coming, for example, from the power system so that it is meaningful in relation to the data coming from the servers. As a result, managers have operated in silos.
This challenge is being addressed through a new generation of data center devices that consolidates data across systems and translates it as necessary for use by the management system.
Additionally, multi-vendor initiatives such as Redfish are creating common communication protocols that will ultimately reduce the need for translation.
As data comes together, so will management teams, presenting a potentially larger challenge than aggregation and translation. Management silos must be broken down and a holistic mindset nurtured to realize the potential of the technology.
Once that happens, the industry will be well positioned to address self-optimization. This will require the development of sophisticated optimization algorithms that do not exist today. This is a complex challenge but one that will be addressed within the context of broader efforts to use data to drive competitive advantage. I don’t expect the data center industry to be a laggard in the use of data to optimize operations.
Conquering these challenges and achieving self-optimization will give us the ability: to maintain optimum conditions across the data center at all times, driving down operating costs; to respond to failures in a way that is transparent to users, improving our value to the business; and to maximize the utilization of resources, reducing capital costs.
As a former CIO I can only say the time can’t come soon enough.
Resistance to Change Will Limit Advancement of Data Center Management
by Mark Monroe, chief technology officer, DLB Associates
While we operate in a very fast-paced environment, the data center industry is, at its core, risk adverse and resistant to change. For that reason, I believe by 2025 we may see industry leaders on the brink of operating self-healing facilities, but the vast majority of colocation and enterprise facilities will be manually operated very similarly to the way they are today.
Full visibility, to me, means that every parameter that makes a difference to the cost of operation or the reliability of services is available to the operators in the timeframe necessary.
For the most part, we haven’t, as an industry, defined the set of parameters that are required for optimization or the ideal frequency with which they should be collected. And, even if we were able to deliver an integrated view of data, most organizations aren’t structured to take advantage of it. I agree with Steve that management silos may be the biggest challenge that we as an industry face in achieving more sophisticated levels of data center management.
But I am not as optimistic that organizations will be successful in breaking down these silos. That will occur only when a new generation of data center managers raised in integrated data center environments assumes leadership roles in a majority of data centers.
Achieving self-optimization and self-healing presents its own set of technical challenges. Self-optimization requires a high degree of intelligence built into data center systems, because we are never optimizing against just one variable. Optimization requires balancing and prioritizing sometimes competing variables, and that introduces a high degree of complexity we are not close to addressing today.
Self-healing — the ability to recognize, isolate and rectify a failure in a system —may be even more of a complex challenge. I contend that developing a self-healing data center is a more difficult engineering challenge than developing a self-driving car. It’s likely that we see autonomous cars on the road by 2025, but I don’t think they will be the norm.
Even though the technical challenges are huge, I don’t think technology will hold back the evolution of data center management. These are difficult but solvable problems.
The reason we won’t see self-optimizing or self-healing data centers by 2025 is because data center operators won’t have enough trust in these systems to rely on them. Remember, these are men and women whose primary role is to ensure the availability of data center services. They are not going to turn that responsibility over to machines until they are absolutely confident those machines can be trusted.
That will happen eventually, but it will take a lot longer than ten years.