Sg.hu – There is a shortage of overnight server admins

Data center operators are most concerned about power consumption and ongoing staffing issues.

The Uptime Institute has released its global data center survey, Global DataCenter Survey 2023 according to material the reliability of the server farms is gradually improving and fewer disruptive outages have been reported. At the same time, admins dare to entrust decisions to algorithms less and less, but personnel and spare parts supply problems continue. The survey covered a total of 879 respondents in several countries, slightly more than half of which are located in North America and Europe.

This was the 13th annual survey of its kind, but the first time operators were asked to identify their top concerns. Among these, improving the energy efficiency of the facilities was in first place, 33 percent of the respondents indicated that they were very concerned about this. The next big concern was finding qualified personnel, while reducing the power consumption of IT equipment was in third place. However, data center managers have also reported that they are having difficulty predicting demand and procuring the equipment needed for growth – a trend that emerged during the pandemic and continues.

The bad news when it comes to energy efficiency is that progress has largely stalled, at least in terms of the most commonly used metric, power usage efficiency (PUE). The annual average PUE value of data centers has been hovering around 1.59 and 1.58 for four to five years. Fortunately, this does not mean the end of the improvement, but rather reflects that all the easy-to-perform conversions have already been learned. Further improving the efficiency of the existing facilities would require significant renovation works, which would be very costly and involve potential downtime.

Not surprisingly, more modern facilities are performing better: 16 percent of this year’s respondents reported an average annual PUE of less than 1.3, mostly in Europe, the United States and Canada. The big question is whether the industry will be able to operate more efficiently in the future. This depends on the resources invested in new technologies and how many older server centers are closed or completely renovated. In Europe, the decision is “facilitated” by high electricity costs, according to the report.

Labor shortages have been a concern among data center operators for more than a decade, exacerbated by the growing demand for capacity. The Uptime Institute reports that while this remains a problem, it’s at least not getting worse at the moment, which they attribute to improvements in areas such as training and staff retention, but the increased use of automation is likely a contributing factor. However, the report warns that the workforce is aging, with older employees making up a significant portion of the workforce. Operators will soon be forced to face the fact that many of them will retire around the same time, taking their knowledge of day-to-day operations with them before training the younger ones.

Half of respondents said they are having trouble finding qualified candidates for jobs, and 35 percent reported that rivals have targeted their employees, double the figure from five years ago. Uptime’s survey also found that the industry is still very male-dominated, with women making up just 8 percent of data center teams and a quarter of respondents saying there are no women in the position at all. This means a lower share even in physically demanding industries such as construction or mining. The report recommends that operators struggling for workers should take steps to include women and other underrepresented groups in the selection process.

In terms of reliability, the Uptime Institute reported a gradual improvement in the frequency and severity of outages. In the report, only 55 percent of operators said they had experienced an outage in the past three years, which is a huge drop from 60 percent in 2022 and even more so from 78 percent in 2020. Another observed trend is that the impact of outages is becoming more and more modest: according to those who answered yes to the previous question, 41 percent of the cases were negligible, and a further 32 percent classified the impact as minimal. Only 4 percent of them reported a critical and another 6 percent a severe fall. According to the Uptime Institute, in the past these two latter categories accounted for about 20 percent of outages. One of the reasons for the decrease may be that, in more and more cases, we can talk about a partial malfunction of the equipment, rather than a complete failure.

More women work in mines and construction sites than in server rooms
However, the costs of outages that do occur remain high, with 54 percent of respondents saying their most recent major, severe, or very severe outage cost more than $100,000, and 16 percent said it cost more than $1 million. Power outages were by far the most frequently cited reason for outages. Other notable findings of the report include the fact that this year, on average, the workload of the company’s own server centers was reduced by less than half, to 48 percent. Of course, this does not mean that the capacity, use or expenses of on-premises data centers will decrease in absolute terms, but rather reflects the increasing use of public clouds.

According to the report, 62 percent of respondents increased data center budgets this year, while 15 percent of workloads are projected to be hosted in the public cloud by 2025. This number may seem low to some, but the Uptime Institute points out that many of those surveyed represent organizations that operate multiple server centers and have many legacy applications, so they are likely to move to the cloud more slowly than companies with smaller IT portfolios.

One of the surprising results of this year’s survey is that nearly three-quarters of the respondents believe that sooner or later AI-based devices will perform certain operations in data centers, and that people may lose their jobs because of this. But in fact, trust in AI has fallen dramatically, with 13 percent fewer people trusting an algorithm to make operational decisions compared to last year, when respondents were asked the same question. As a result of the increased media appearances related to artificial intelligence, there has been a greater awareness of the flaws and limitations of the current models, as well as the knowledge of the tools based on them. But even if AI models become reliable enough to take over certain data center roles, the short-term impact would be very minimal, notes the Uptime Institute, as it would only fill labor shortages and reduce the need for additional hires, not replace those already there.