High-Performance Computing Helps University Researchers Do the Impossible

By Paul Abram, Data Center Practice Leader, and Nick Turk, Solution Architect

Today’s colleges and universities are vying for more than just students.  They’re competing for research grants and notoriety as well – and their success, in many cases, is largely dependent on the robustness of their IT systems.  With the sheer volume of data created and accessed by each school’s research teams alone, higher ed CIOs have been tasked with finding, implementing and managing computing systems optimized for data analytics applications.

To compete in new and innovative ways, university research teams need sophisticated analytics engines to discover, interpret and analyze the unimaginable quantities of data now available to them from both internal and external sources – and to do so in shorter time spans than ever before.  But, according to the fifth annual Logicalis Global CIO Survey, a poll of 890 CIOs in 23 countries, complexity remains the largest barrier for CIOs around the world in getting the most from business analytics.

One of our clients, a Big 10 university, was in this exact position.  Before Logicalis was brought in, they had planned a very large-scale deployment of white box X86 servers – a monstrosity of an environment. Every server would have to be racked and stacked, networked and cabled.  They would have needed a lot of switches, and of course, on the soft side, additional power, cooling and maintenance. We showed them, however, that an IBM Power Systems solution could provide more horsepower in a significantly less complex design.

In fact, high-performance compute solutions such as IBM’s Power Systems platforms can help researchers get more work done faster, with less equipment, and oftentimes more affordably than simply scaling out a complex cluster of x86 servers

The key is the efficiency and scalability of the platform. With Power, CIOs in higher education can deliver at least twice the amount of work per core than they can with an x86 platform – and that’s the bare minimum. With many workloads, IBM Power Systems can provide 3X guaranteed performance improvements, and in some cases, 50X to 100X improvements.  The thing university CIOs – as well as CIOs in vertical markets from healthcare and finance to retail, government and manufacturing – need to know about Power Systems solutions is that it really is possible to do today what was impossible just a couple of years ago.

Is Power the Right Option?

To determine if a Power solution is right for a university’s high-performance computing needs, it’s critical to fully understand analytics requirements, data sources, the movement of data throughout the university’s environment, as well as the school’s overall IT resources.

Understand Your Analytics Requirements

Before making a new investment or a change to your data analytics compute environment, it’s essential to fully understand what you are trying to accomplish. For example, are you simply trying to give the university’s research department additional data-crunching capabilities, or do you want to extend those capabilities to all departments campus-wide? Some universities may need high-performance data analytics for examining medical use cases in one department as well as seismic or weather modeling in other departments. A general-purpose analytics engine can be used to meet your school’s varied needs rather than implementing siloed strategies for each department, so looking at the combined requests of each department and implementing a solution that works for all may be the most cost-effective route to success.

Understand Your Data Sources

Know where your data sources originate as well as where they could originate and all the ways that data may need to be used and manipulated before implementing a solution. Imaging, seismic, space and weather data may be gathered once, then used in different ways by different departments. One research group may want to analyze the metadata, while another may be interested in the location and machines that generated the data, and still another may want to examine the images themselves pixel by pixel. Knowing what kind of data is being gathered and which departments need access to that data will help you determine the platform that can deliver the best performance in the most elegant way possible.

Understand the Movement of Your Data

In research analytics, there is a huge amount of research data being manipulated and, often, shared within a university. Before implementing a compute solution, therefore, it’s important to examine how and when that data will be stored, shared and moved throughout the system. The goal is to move these tremendously large data sets as infrequently as possible and to store them only once without the need for duplication. As a result, it’s important to think through the university’s networking and data storage requirements very carefully before deciding on a strategy.

Understand Your IT Resource Needs

To provide a common analytics infrastructure that all the university’s departments can tap into, start by analyzing your school’s compute, networking, storage and security requirements campus-wide, then design an infrastructure that can support all those needs. Even if your investigation began with a particular research department’s request, it won’t be long before the university’s business leaders and developers will be asking for analytics capabilities as well.  Therefore, you’re going to need a common infrastructure that is scalable, manageable and sustainable – and one that can continue to grow affordably.  While both IBM Power Systems and x86-based solutions can offer the compute power you need, the Power Systems architecture delivers a unique ability to scale both affordably and securely.

Want to learn more? Read about some of the ways Logicalis is helping CIOs in higher education use technology to solve some of their biggest challenges. A good place to start is with a professional assessment of your university’s existing infrastructure. Find out how Logicalis, an IBM Platinum Business Partner, can help you transform your university into a digitally enabled enterprise, unlocking the potential and value of the data you collect and store.

Leave a comment

Your email address will not be published. Required fields are marked *

Shares