<div class="page" title="Page 1">
<div class="section">
<div class="layoutArea">
<div class="column">
<p><strong>Australia is at a tipping point for data-intensive research <span style="font-weight: 400;">—</span> here are the key challenges that we need to overcome to safeguard the sector’s future.</strong></p>
<p>Accelerated and specialised computing technologies have become essential to research today, with GPU computing and deep learning being behind many of the world’s <span><a href="https://www.nature.com/articles/s42256-022-00463-x">most recent and transformative scientific breakthroughs</a></span>.</p>
<p>Despite Australia being <span><a href="https://www.science.org.au/curious/policy-features/science-australia">well-known for its scientific contributions</a></span>, our current infrastructure is beginning to fall behind our international peers due to the reliance on outdated technologies, impacting both the functionality and performance of our research sector. While researchers have access to more data than ever before, our current infrastructure is often unable to support the in-depth analysis required to find the meaning hidden within this data.</p>
<p>This is compounded by <span><a href="https://www.science.org.au/curious/policy-features/science-australia">recent shifts in Australia’s research &amp; development funding</a></span> structure prioritising work with practical and/or commercial outcomes at the expense of discovery and fundamental research &#8211; the areas which most often lead to world-changing discoveries.</p>
<p>The Australian research sector is at a tipping point, and in order to safeguard the future of this industry <strong><span style="font-weight: 400;">—</span></strong> and remain competitive with our international peers <strong><span style="font-weight: 400;">—</span></strong> it is imperative that our researchers have access to the leading technology that will enable them to unlock the full potential of their data-intensive research.</p>
<p>Firmus’ Supercloud provides a solution to some of the challenges this industry is facing, working with researchers to ensure that Australia remains at the forefront of research and development.</p>
<h2><strong>Challenge 1: Availability of the latest GPUs</strong></h2>
<p>Much of Australia’s research is dependent on the availability of Graphical Processing Units, or GPUs, which enable researchers to perform specialised computations in parallel and save weeks <strong><span style="font-weight: 400;">—</span></strong> if not months <strong><span style="font-weight: 400;">—</span></strong> of time in data processing.</p>
<p>GPU technology is continually being improved, but the rate of improvements makes it difficult for researchers to keep up, as they are unable to obtain the latest generation offerings due to availability or reduced funding. While there are plans to add more GPU capacity to Australia’s research infrastructure, this is not an easy task and may take years to implement.</p>
<p>By partnering with NVIDIA, Supercloud gives researchers in-country access to the latest generation of GPUs as they are released by NVIDIA, enabling researchers with high-security requirements access to this technology. These GPUs are hosted virtually on Supercloud’s servers and are always kept up to date, so researchers can be sure that their data is always being efficiently processed.</p>
<h2><strong>Challenge 2 &#8211; Maintaining or optimising code for newer GPUs</strong></h2>
<p>While access to the latest GPU hardware is integral for data-intensive research, the speed and efficiency at which this hardware can operate are dependent on updated software. It is therefore crucial that GPU codes and models are kept up to date to ensure optimum performance and function. <span> </span></p>
<p>This can be a significant challenge for researchers, as refactoring code takes both time and technical expertise and designing architecture and software for modern GPUs is not straightforward. This is an even greater problem for research projects that are making the jump from CPU to GPU, as this requires a fundamental restructuring of their existing systems.</p>
<p>For example, the latest NVIDIA GPUs support Multi-Instance GPU which allows for more efficient utilisation of GPUs through secure partitioning for multiple applications. This is a wonderful technology with an infinite amount of applications, but it cannot be used without the libraries, software, and knowledge needed to expose its capabilities.</p>
<p>Supercloud’s service includes access to experienced solutions architects who can advise researchers on code optimisation for the latest GPUs, alongside integrated profiling and development tools for GPU code optimisation, such as Nsight.</p>
<h2><strong>Challenge 3: Cooling</strong></h2>
<p>The increase in cooling requirements to run and maintain the accelerated compute that is integral to research continues to grow, with cooling systems accounting for <span><a href="https://www.nature.com/articles/d41586-018-06610-y">up to 40% of a data centre&#8217;s total energy usage.</a></span></p>
<p>The traditional air-cooling approaches currently used by Australia’s data centres are insufficient and outdated for the demands of data-intensive research and the latest-generation GPUs. The inefficiencies in these cooling approaches are in turn having a significant environmental impact, with the high electricity consumption of legacy data centres being <span><a href="https://www.nature.com/articles/d41586-018-06610-y">one of the leading causes of CO2 emissions</a></span> within the tech industry.</p>
<p>Unfortunately, with hundreds of millions of dollars already invested in legacy data centres, it is often not financially viable for these data centres to move to more energy-efficient immersion cooling.</p>
<p>Supercloud’s data centres are designed and purpose-built for immersion cooling, with the immersive fluid being 1000x more efficient than air at removing heat, significantly lowering energy consumption and carbon emissions. The Supercloud data centre is powered by Tasmania&#8217;s renewable energy supply &#8211; so the energy that is consumed is nearly carbon-free.</p>
<h2><strong>Challenge 4: Lack of cost-effective access to commercial-grade infrastructure-as-a-service</strong></h2>
<p>Commercial service-level agreements and security accreditations are not available through Australia’s current research infrastructure, and while small or low-risk research projects may be able to tolerate sub-optimal system reliability and security, large, sensitive, or high-risk research programs cannot.</p>
<p>While hyperscale commercial providers do offer reliable and secure solutions for researchers, projects can often suffer from scalability issues due to the limited capacity or increasing costs of the provider.</p>
<p>Supercloud offers the same service-level agreements as the big cloud providers at approximately half the cost, with the inclusion of ongoing access to senior solutions architects.</p>
<p><span> </span>The Australian research sector is at a tipping point, with researchers across the country challenged by the availability and accessibility of latest-generation GPU hardware and software, the significant environmental impact of legacy data centres, and a lack of cost-effective access to commercial-grade infrastructure-as-a-service.<span> </span></p>
<p><em>If you’re a researcher facing these challenges and looking for a solution, feel free to contact <a href="https://supercloud.co/">Supercloud</a> to learn how we can help.</em></p>
</div>
</div>
</div>
</div>


DSC07404-scaled

Challenges for e-Research in Australia

<div class="page" title="Page 1">
<div class="section">
<div class="layoutArea">
<div class="column">
<p><span>We&#8217;re pleased to announce that the Supercloud Image Library has undergone a major upgrade. The new library is cleaner than it used to be with improved image naming and better automation to keep the images up to date. </span></p>
<p><span>The library contains Supercloud-specific virtual machine images and base Linux and Windows images that are configured and optimised for deployment onto the Supercloud compute platform. </span></p>
<p><span>The Supercloud Workstation Images (SWIs) have been rebranded and now have consistent names. You can find all of the Supercloud images in the dashboard under Compute -&gt; Images. </span></p>
<p><span>The details of each image can be seen by clicking on the image name in the dashboard. Here you will see the image specifications and other information that you might need. Images can be launched from the library using the “Launch” button. </span></p>
<p><span>The Supercloud Deep Learning Image (Deep Learning SWI) is our flagship image for high-performance GPU-based machine learning workloads in either research or production environments. It includes Tensorflow and Pytorch for deep learning, and a suite of popular tools, such as Scikit-Learn and Jupyter, for data science and data preparation tasks. Launch this image to begin developing machine learning models without worrying about drivers, frameworks, and version compatibility issues. </span></p>
<p><span>There is also a GPU base image (GPU Base SWI) in case you need the GPUs to be configured, but don’t require specific pre-installed frameworks. This image is ready to go with all the drivers and configurations needed to start using GPU compute on accelerated SCE instances. The GPU Base SWI is a good choice for running NGC docker images or for installing GPU-heavy applications for 3D rendering or photogrammetry. </span></p>
<p><span>The image library also contains a selection of upstream Linux images. These images are automatically updated as new versions are released, so you can be sure that you always have access to the latest versions. Older versions are also available if you need them. </span></p>
<p><span>The Supercloud Windows images can be launched and accessed through the dashboard, or, once running, can be accessed using any remote desktop (RDP) client for Windows. The Windows SWIs running on Supercloud SCE offer reliable environments for Windows-based applications. There are both desktop and Windows Server images available, which have been configured to support GPU-based instance types. These images offer a cost-effective solution for running high-performance GPU workloads for Windows but are also perfect for running general-purpose Windows applications. </span></p>
</div>
</div>
</div>
</div>
<div class="page" title="Page 2">
<div class="section">
<div class="layoutArea">
<div class="column">
<p><span>There are tutorials on how to use the images in the library, which step you through the process of launching instances and logging in. These tutorials are available to all Supercloud customers, and our team is here to support you with the image library or anything related to Supercloud. </span></p>
<p><span>We’d love to hear from you and help you get started! </span></p>
</div>
</div>
</div>
</div>


Sign up

Sign up

Sign up

Sign up

Sign up

Challenges for e-Research in Australia

[Product Update] New Supercloud Image Library

Filter

Step into a sustainable future.
Get started today.

Get in touch

Sign up

Sign up

Sign up

Sign up

Sign up

Challenges for e-Research in Australia

[Product Update] New Supercloud Image Library

Filter

Step into a sustainable future.Get started today.

Get in touch

Step into a sustainable future.
Get started today.