The University has chosen to rely on E4’s experience in the creation of very high performance infrastructures, using the E4 USTI solution for artificial intelligence applications

E4 Computer Engineering, one of the leading Italian companies for high-performance solutions for HPC, artificial intelligence and containerization, is the partner chosen by the University of Pisa to guarantee a solid and performing IT experience for its researchers who develop artificial intelligence models for different application areas.

Since the end of 2016, the University of Pisa has created a Green Data Center, based on state of the art technologies, designed to be the infrastructural engine and enabler of digital innovation initiatives, that need Big Data processing, cutting-edge technologies and expertise in HPC and Artificial Intelligence fields.

Nowadays, the enormous amounts of data available to companies have the potential to open up new business opportunities, but it is important to manage them efficiently, through increasingly performing and flexible infrastructures, especially with regard to the storage, which must offer speed, agility and durability.

Machine learning and deep learning algorithms, in particular, tend to make more frequent I/O requests to a larger number of small files, compared to traditional HPC applications, so, having a fast and efficient large-scale data processing becomes essential.

To face this challenge, the University of Pisa has therefore decided to rely on E4’s experience in high-performance infrastructures, implementing the USTI (Ultrafast Storage, Totally Integrated) solution within its Green Data Center – designed entirely by E4 engineers – powerful, scalable and reliable, 4-node, based on Excelero’s software-defined block storage NVMesh and able to take advantage of the near-zero access latency of NVMe SSDs.

Using USTI as the foundation of its own accelerated infrastructure, the University of Pisa was able to eliminate storage performance difficulties that could penalize the overall result, allowing the use of computational resources mostly aimed at artificial intelligence applications, where E4 USTI has proven to be able to provide significantly superior performance (about 1.8x) compared to traditional software, using the same underlying technology.

These are two examples of artificial intelligence applications the University of Pisa has worked on, exploiting USTI infrastructure:

• CoVID19 Drug Repurposing – For this application, deep learning models for graphs have been used, to identify which could be the more suitable drug, among the ones already on the market, to be used in anti-COVID19 therapies, with the intent of creating a tool that can speed up the preliminary screening process in therapies.

• Autoencoding Music Generation – In this application a model, called Calliope, was built in, able to learn by representing various music styles in an organized way, within a neural network that can then be used to generate new music tracks.

“The Green Data Center of the University of Pisa is implementing an innovative computing and storage architecture, to support new scientific computing workloads (AI, ML / DL, Data Analytics, Genomics…) “, explains Maurizio Davini, CTO of the University of Pisa. “USTI allowed us to complete at best GPU Computing infrastructure and has proven to be an ideal solution for ease of implementation and for the exceptional performance it offers to our researchers.”

“We are proud of the project we have created for the University of Pisa, which confirms USTI as the ideal solution for today’s infrastructures and applications, which need to have a latency close to zero” says Cosimo Gianfreda, CTO of E4 Computer Engineering. “With USTI, it is possible to maximize the performance of the entire infrastructure, using acceleration technologies in combination, such as NVMe disks and NVMesh software, thus optimizing the use of all available resources.

To know more, click here