Facility Design Services
Your facility may require updates or modifications in order to successfully support your new systems. Factors such as humidity, airflow, power draw and accessibility all need to be taken into account for every installation, and your facility may be incapable of supporting the higher power and cooling requirements engendered by more densely racked HPC systems. Aspen Systems knows data center environments and can help you upgrade your existing facility, or even design a new one. When Aspen Systems helps design or upgrade your facility, you know that you’re getting the best equipment, engineering, and integration available. We also "future proof" our designs in order to provide the ability to painlessly grow your facility in the future. Aspen Systems has successfully designed new facilities and upgraded existing computer facilities for more than 20 years – let our next success be your success.
Existing computer facilities represent a significant investment. Most facilities are designed with a 10 to 15 year life cycle in mind, meaning that you would expect that the power and cooling infrastructure to adequately support your computing needs for that period with only simple maintenance. So why do many organizations experience cooling or power issues in their current facilities? The problem isn't that the original components or implementation was flawed, it is that modern computers have evolved so quickly, far outstripping the maximum power and cooling capabilities that most existing compute facilities were designed to provide.
Modern clusters and computing equipment become more and more dense with every generation. While new systems are much more power efficient for the processing power you gain, they're also being installed in less space than ever before. Chassis formats such as blade servers, "2 in 1U", or "4 in 2U" systems and other highly compact system designs save space and make it easier to get more processing power in a smaller footprint. Modern multi-core processors and denser memory formats are now also common design selections. Space has become a very valuable resource, and modern HPC and enterprise computing designs reflect this fact.
However, most data centers were never designed with adequate power or cooling for today's denser solutions. Recently, the American Society for Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) changed their data center maximum ambient temperature recommendation from 77° Fahrenheit to 80.6° Fahrenheit. In part, this was done to lower ever-rising cooling costs, but it also reflects the fact that modern architectures are designed for hotter environments, and that facilities are having more and more difficulty maintaining lower temperatures.
Most data centers built in the last two decades were designed to provide adequate power and cooling in the facility to service 3 to 4 kilowatts (kw) of power and waste heat dissipation per rack. Modern cluster implementations almost always far exceed this density. Aspen has measured many different HPC systems to arrive at a "real world" load estimate for many of the standard compute nodes used in HPC today. Using our standard 1U server loaded power usage (322 watts), a rack of 40 1U systems under load will produce a 12.88 kw heat load, over 3 times the maximum amount of heat most facilities were designed to handle in a single rack. If your cluster design is heavily loaded or your code is more compute intensive, this load could even be higher.
Using higher density systems such as "2 in 1U", "4 in 2U" systems, or blade servers, the heat load can be more than 6 times the amount of heat the average facility can reliably dissipate from a single rack. Most HPC designers strive to deploy clusters that are as heavily utilized and densely racked as possible. The most densely racked solutions can produce over 27 kw of waste heat in a single rack. High density racks often overwhelm the cooling capacity in a single area of a traditional computer room, causing hot spots. If you are deploying a large HPC cluster, the entire facility temperature can be raised to unacceptable levels because your current facility can not handle the additional heat load generated by your new systems.
