Advanced Computing in the Age of AI | Saturday, April 20, 2024

What’s Stopping Datacenter Progress? 

Technologies such as virtualization, cloud computing, big data, and converged infrastructures are the standouts of modern datacenter design. Yet most datacenters in operation today are unable to implement these best practice principles. 

Technologies such as virtualization, cloud computing, big data, and converged infrastructures are the standouts of modern datacenter design. Yet most datacenters in operation today are unable to implement these best practice principles. A recent report from power management company Eaton offers some perspective on this all-too-common scenario.

According to a recent white paper by John Collins, Global Segment Manager, Data Centers, Eaton Corporation, businesses that adopt these cutting-edge technologies experience decreased IT overhead, increased business agility, and find more ways to create a profit. Yet, these benefits remain out of reach to a large swath of datacenter operators. The reasons, as Collins explains, are multifold.

For one, older datacenters struggle with inadequate power density. The datacenters built about eight years ago were designed to support 3 kW per server rack, or 100 watts per square foot. The more recently built sites can support 6 kW, or 200 watts per square foot. However, datacenters today require as much as 15 kW per rack or 600 watts per square foot.

Many datacenters also lack power and cooling capacity. Right now they have a tough time supporting the present requirements but as companies start to adopt cloud computing and big data technologies in greater numbers, the struggle will only intensify. As for cooling capacity, businesses often plan for neatly aligned rows of server racks. As new equipment is added over time, businesses find themselves with insufficient cooling capacity in these high-density areas of their datacenters.

Not only do many company datacenters lack power and cooling capacity, but they have inadequate redundancy as well. Over time, as businesses rapidly raise their power density, they often neglect to increase their computing facilities’ uninterruptible power system (UPS) and cooling capacity at the same rate. Down the line, these will end up offering no redundancy at all.

Many datacenters also have an absence of rack deployment flexibility. In order for these datacenters to accommodate big data or cloud-based approaches, servers either need to be added or moved around. This is a problem because the racks and cooling systems that most companies currently use limit their positioning options.

While leveraging the latest and greatest datacenter technologies sometimes seems to create more problems than solutions, the report offers some tips and techniques for an easier upgrade path.

For example, when acquiring UPS hardware, businesses need to be aware of models that provide documented energy efficiency and flexible thermal management. They should also have flexible placement options and concurrent maintenance capabilities. Other important characteristics are modular scalability, serviceability, and redundancy.

According to Collins, intelligent, compact power distribution is best for keeping efficiency and density high and power distribution low.

When considering server enclosures, the hallmarks of good design are flexibility and efficiency. They best ones will accommodate multiple deployment and rack options while having minimal air leakage according to Collins' report.

The last point that Collins makes on this subject is that businesses would be wise to invest in software that provides unified administrative control, can perform live virtual machine migrations, aggregate power protection and distribution device information, and protect workloads during power outages.

By upgrading their datacenters with these specifications, businesses will be able exploit the competitive advantage of modern design approaches and will be in a better position to leverage future IT developments.

EnterpriseAI