The demand for increasingly powerful computing has led to a cascade of demand across many data centers. As technology has advanced and allowed computers to do more, businesses and consumers seem to be always ready to fill the computing void. Even the average consumer is aware of the ongoing explosion of digital utilities of every type, and the dizzying capabilities of digital technology in general.
A resurgence in HPC
This has given rise to more and more enterprises looking into high performance computing (HPC). In decades past, HPC was favored by mathematicians and research scientists who required computing power far beyond the needs of anything commercially available in order to execute complex mathematical calculations.
Today, the demand for HPC is far greater. In addition to the realm of research science (which continues to demand increasing computing power), there are financial institutions doing risk modelling, governments calculating the impact of changing demographics and aerospace companies analyzing the flight capabilities of aircraft and spacecraft. Imaging systems alone—which covers the demands of everyone from health care to the movie industry—can require mind-boggling levels of computing power.
Because of the high cost of securing, deploying and operating a supercomputer and its proprietary software, this technology is still out of reach for most businesses.
Integration Over Investment
While it may sound like something from a recent action movie or TV series, the integration of multiple computers into interconnected clusters in order to facilitate HPC capabilities actually is, as they say these days, “a thing.” In this scenario, computers with compute cores (processors), graphical processing units (GPUs) and memory are organized in multiple nodes, effecting a viable supercomputing system.
These clusters, consisting of inexpensive computers running commercially-available software, are substantially more affordable than a supercomputer—but they still represent a significant investment for most organizations.
Turning to the Cloud
With the growing popularity of the cloud and the increasing and diverse utility it offers, some businesses would like to achieve HPC access without breaking their IT budgets. They are turning to the public cloud services offered by tech giants like Google, Microsoft, Amazon and IBM. Engaging these services allows them access to HPC capabilities without all of the investment in hardware infrastructure.
These cloud-based utilities have also served to significantly level the playing field, giving smaller companies the ability to break out with HPC alongside large enterprises. In some cases, organizations that had successfully adopted integrating clusters later found that user demand was exceeding compute resources and capacity. Many of these companies have been able to meet the demand through the cloud, which affords a flexible, on-demand HPC environment.
This approach has proven to be cost-effective for many organizations. The available platforms also have user-friendly, efficient features for scheduling and streamlining workflow. Now, more and more businesses are dependent upon the collection, analysis and distribution of data. They see a need to move to the next level of computational power, and they will likely turn to the cloud for their HPC needs.
Data Center Considerations
For those organizations that are implementing HPC installations on their own, it’s important to address what will happen in their data centers: two of the biggest challenges being increased power density and increased heat loads.
Managing Increased Heat Loads
Industry-leading Smart PDUs or Switched PDUs are providing the solutions that power today’s higher performance computing applications, with 3-phase power and load balancing capabilities.
- A balanced load means less heat: the first point of value that a Smart PDU presents is arguably the best line of defence – not having to worry about heat in the first place. That is the operating philosophy associated with balanced electrical loads in three phase applications. A balanced load is a happy load, a more efficient load, and, most importantly, a cooler load.
- Alternating Phase Technology: for most data center managers, the rear of the cabinet is off-limits. In order to clear a path for airflow, cable management must be as well-ordered as a circuit board. ‘Alternating-Phase’ technology offers a unique approach to the placement of outlets on Smart or Switched PDUs. Traditional PDUs have the X, Y, and Z phases grouped together: X,X,X; Y,Y,Y; Z,Z,Z. In an Alternating Phase configuration, those phases are grouped X,Y,Z; X,Y,Z; X,Y,Z. Plugging into a different phase is then a matter of inches, not a long power cord run to the bottom of the rack, significantly improving cable management.
- Use metering to go with the ‘flow: built-in metering and environmental monitoring can alert you to trouble areas in your data center, and help you understand the power consumption cycles of you brand new HPC cluster. Not only will a Smart PDU help you avoid an unfavourable overcurrent condition, it can give you immediate feedback on where cooling may be needed.
Increased Power Density
One of the key challenges facing data center managers in the fast-moving world of HPC is being responsible for specifying a rack PDU that can accommodate future equipment changes and plug types. High Density Outlet Technology (HDOT) PDU solutions are available that provide the most outlets in a 42U intelligent rack PDU, to support all of those extra power cords. Configurable PDUs like the HDOT Cx, feature an outlet that acts as a hybrid of the standard C13 and C19 outlets, accommodating both C20 and C14 plug types. This 2-in-1 outlet offers flexibility and has the ability to adapt to different rack configurations – enabling one PDU to support many changing rack configurations.
Whether your business is looking to implement HPC installations via the public cloud or internally in your own data center, a flexible, on-demand environment is key.
About the Author
Rebecca Gilstrap is Director of Strategy for Legrand’s Data, Power and Control Division. In this role she leads cross functional teams to set, prioritize and achieve strategic growth initiatives. Rebecca has spent the last decade ardently optimizing data centers and mitigating IT infrastructure outages. With a career rooted in security compliance and business continuity, she has worked across organizational boundaries to highlight critical IT dependencies. She has managed international technology divestiture projects, built colocation businesses from the ground up and created multi-tenant cloud service platforms. She has a passion for high performance buildings and utilizing technology to create better experiences and maximized ROIs. Rebecca has held DOD clearance as well as Sensitive Compartmented Information Access. She holds a B.A. from the University of South Carolina.