CERN openlab and IBM are organizing a Technical Deep Dive workshop at the IBM Research Lab in Rüschlikon.
The goal of the workshop is not to exchange presentations, but to identify the right scope for research collaboration within two to four proposed areas of interest and establish a concrete set of actions to set up a broad, forward-looking research partnership between CERN and IBM Research around the most pressing IT challenges that CERN, HEP and the LHC Experiments will face in the near future.
The four possible areas for discussion are:
Accelerating computing:
In this session we will dive into disruptive systems and hardware technologies conceived to accelerate AI/ML algorithms for data reduction/information extraction in environments where data streaming is a bottleneck. In particular, we will detail how in-memory computing can drastically reduce the algorithmic complexity of machine learning/deep learning primitives, and identify use cases that could have a direct impact to CERN’s data streaming challenges (e.g., temporal correlation detection, deep learning training and inferencing).
Data storage:
In this session we aim to identify how a number of solutions developed at IBM Research can be leveraged and/or shaped to tackle CERN’s long-term data storage challenge and IT infrastructure scaling hurdle. Firstly, we propose to review the latest innovations in tape technology hardware and software to address data growth and integration in common file systems (e.g., LTFS-DM), and the adoption of AI-driven techniques to predict access to data and stage it to/from tape. Secondly, we propose to focus on NVM storage systems, and storage services designed to most efficiently leverage NVM, such the use of NVM instead of DRAM in memcache, and use of ephemeral NVM storage to accelerate Spark and serverless workloads with Crail.
https://www.zurich.ibm.com/cloudstorage/software.html
https://www.zurich.ibm.com/sto/tape/arealdensity.html
Quantum Technology, Algorithms and Applications
In this session, we will first set the scene with an introduction to IBM’s quantum computers and quantum programming/simulation framework (Qiskit, Aqua, Terra). We will then deep-dive into quantum algorithms for accelerating machine learning, for optimization and combinatorial searches, and for running quantum simulations. The goal is to identify key applications of quantum for CERN, with a focus on those having quantum algorithms that could leverage available hardware in the near-term to demonstrate a clear quantum advantage.
Knowledge Management and Discovery:
In this session we aim to define how IBM’s Cognitive Discovery assets could be leveraged to efficiently manage and better value CERN’s scientific knowledge base by making it more available to scientists. Applying a scalable PDF ingestion system with knowledge extraction and representation algorithms, combined with a deep search engine will enable scientists to be a click away from surfacing relevant information in their context from digital libraries or scientific repositories.
______________________________________
The proposed agenda depends on the interests of the participants, topics can be added or removed as needed.