The next generation of HPC networks: how to adapt to keep pace with growing bandwidth demands
It’s an exciting time to be involved in high-performance computing (HPC) in Europe. The EC is investing up to a billion euros in a new facility – the European Open Science Cloud (EOSC) – which will help boost Europe to third position globally in the field of high-performance computing by 2022. The revolutionary HPC cluster, which will be built on all-European technologies, will drive greater collaboration between HPC facilities, research institutes, and private organisations across Europe, presenting a wealth of new research-based and commercial opportunities.
As research partnerships and collaborations and associated opportunities proliferate, however, so does the data being generated by research partners, from CERN to European large telescope arrays.
Can your network connections keep pace?
HPCs are in the business of providing massive computing capabilities, with national providers such as GENCI in France delivering 2 flop/s of calculation power (equal to two million, billion operations per second). While your core business requires you to invest in the infrastructure required for computing at these astronomical speeds, however, network connectivity is often a barrier to rapid data transfer between HPCs, your partners, and your clients.
This is a significant challenge, as massively growing data transport requirements put National Research and Education Networks (NRENs) and other research networks under pressure. Additionally, the rapid growth in HPC facilities and partnerships requires rapid connectivity to new sites and the ability to scale connections on-demand – features that traditional NREN managed services are typically unable to provide.
In the face of these challenges, many HPCs are looking for ways to increase control over their network connections, while also increasing the efficiency and cost-effectiveness of data transport between organisations and locations.
The question is, how can HPCs multiply available bandwidth, scale their networks on demand, and deploy new connections to partner locations quickly, while still focusing on your core business of delivering vast processing power to the research community and enterprise organisations?
Future-proofing HPC networks with alien wavelengths
One way HPCs can keep pace with growing bandwidth demands is by deploying alien wavelengths in your network. These can help you multiply available bandwidth on existing photonic line systems by up to 250G on 50GHz systems, or up to 400G on flexible grid, or 100GHz systems – allowing you to deliver vast datasets to your partners and clients much more quickly.
In addition to these benefits, alien wavelengths can help you monitor, manage and scale your network connections more easily. You can also deploy new connections programmatically and scale up available bandwidth in real time, allowing you to connect with new partners and clients in hours, rather than weeks or months. You can also partner with other HPCs at remote locations and deliver new, revenue-generating services to private organisations.
Delivering HPC data over dark fibre
Another way to keep pace with rapidly growing data transport requirements, while also lowering networking costs, is to create your own network connections over dark fibre on your NREN or other provider network. Rolling out a dark fibre or self-provisioned network need not be difficult: optical transmission and packet technologies are now mature, and meet all the technical requirements with very little need for specialist skills.
In addition, the operational benefits of lighting dark fibre are compelling. With full control over network equipment and the ability to converge services from layers 0 to 2, dark fibre networks offer major cost savings per bit compared to managed services alternatives. Additionally, they offer security that surpasses shared network connections, excellent Class of Service predictability, always-on reliability, and rapid service turn-up and scaling.
With such clear benefits compared to traditional network managed services, is it time for your HPC to light dark fibre connections or adopt alien wavelengths? Look out for my next HPC connectivity blog, where I'll discuss more great reasons to re-think your network strategy today - or reach out to discuss your options today.
Find out more . . .
As the HPC landscape in Europe evolves at pace, new networking technologies, such as alien wavelengths and dark-fibre networks, are needed to support fast, cost-effective data interchange between partners and customers. To find out more about next-generation HPC network technologies and how they can help you future-proof your operations and maximize commercial opportunities, watch out for the second blog in the series.