How GRAID servers are helping to conduct world-leading financial modelling research

Conversation 

In the world of quantitative finance, a leading UK institute needed cutting-edge technology capable of driving their research forward. Using advanced AI and Machine Learning for financial modelling, they required a powerful GPU server and scalable storage, capable of seamless communication in order to prevent workflow bottlenecks. 

Collaboration 

To meet the institute's unique requirements, we embarked on a collaborative journey, seeking to understand the use-case for these systems, and identifying how we could tailor this hardware to their unique requirements.  

The first milestone involved delivering a state-of-the-art Nvidia L40-based GPU server. Engineered to accelerate workloads, this powerhouse was complemented by a high-speed NVMe 4.0 SSD-based GRAID storage server, boasting an impressive 178TB capacity. To prevent any potential data loss, we utilised RAID 6, which provides fault tolerance for up to two drives in every pool. Both solutions were designed to be fully scalable, ensuring adaptability for future needs. 

The second, was to ensure fluid communication between the servers, so we settled on implementing a high-speed 100GBe Backbone network, to avoid any bottlenecks. This infrastructure not only delivers impressive bandwidth and speed but is also easily scalable for future expansion. 

The institute also needed to ensure that their research would be kept completely secure. As such, we made sure to include automounted home directories with centralised group management, providing precise access permissions. We established a secure identity management and access control domain with Kerberos authentication, limiting access to authorised personnel only. 

Lastly, we migrated their existing GPU and CPU nodes over to their new centralised identity provider and integrated a SLURM Workload Manager to streamline resource utilisation, helping the institute to conduct their research more efficiently.  

  

Confidence 

The institute now has a high-performance GPU server and a lightning-fast storage server, which can communicate with each other without bottlenecks. This means they can conduct their world-leading financial modelling confidently and without limitations. Their solution is also completely scalable, so it can easily be expanded as their research progresses.