Systems Research in Tokyo

The Systems Group in IBM Research - Tokyo (TRL) has long research experience on systems software for more than 20 years - including Java, middleware, and high-performance computing (HPC).

The team has been contributing to:

• IBM Java JIT compilers
• IBM Java VM improvements
• HPC application acceleration
• Big data applications
• Blockchain
• Optimizations for POWER
• Optimizations for System z
• Open source contributions
• and others ...

A list of selected publications by the members can be found below. Recently, the team has also been expanding the systems software expertise to Hybrid Cloud and AI.

Research on Hybrid Cloud Infrastructure

Hybrid Cloud is considered as an IT infrastructure to connect more than two clouds --- such as AWS, Azure, Google, IBM, private, etc. The infrastructure for it should support various constraints of data, security, and economics. In Tokyo, the team is working for the analysis and optimization of various cloud workloads. For example:

• The team provided diagnoses, insights and tools to understand container performance [IEEE CLOUD 2020]
• The team optimized application performance based on system and application metrics [IEEE MASCOTS 2020]

Research on AI Chip Software

In the AI Hardware Center, IBM is developing various AI hardware. Tokyo team is working for the development of a software stack named DeepTools for the RaPiD digital AI core, especially for compiler and simulator as shown below.

Open Source Participation

The team is leveraging and contributing to open source software. Some examples are:

• Cloud - Kubernetes, Apache Spark*, OpenJDK*, Death Star Bench, SPDK, GATK
• AI - ONNX, MLIR, LLVM**, TensorFlow, Apache Arrow
• Others - Hyperledger Fabric, Qiskit

For the projects marked as "*", the team has committers.

Selected publications can be found below, older publications can be found here.