Jin Li, Principal Researcher & Research Manager
Cloud Computing and Storage
MSR Technologies
Email: jinl@microsoft.com

OneNet: Distributed Functional Computing

Dr. Li's latest passion is OneNet. OneNet is a high performance distributed, unified functional computing platform written in F#. It is designed to unify batch data processing, streaming data processing, real-time machine learning and real-time query processing, all under one platform with shared in-memory instantiated data with concurrent access support. OneNet is extensible for system builders (e.g., adding support to Azure is less than 250 lines of codes), natively supports computation across multiple clusters, and can mix and match specialized device compute (e.g., GPGPU/FPGA/Smart SSD) with CPU based cluster computing. OneNet can speed up the engineering (including development, debugging and deployment) of building a complex distributed system by >3x, transforms large-scale data analysis, and enable a cost competitive cloud computing service. OneNet achieves these via distributed functional computing, so that 1) programmers can easily build a highly complex distributed system by expressing the system as nested evaluation of functions, and 2) the program written can be native executed remotely, which enables fine-grain control of remotely executed program extensively sharing of in-memory data within and across jobs. From his personal experience in developping, deploying and debugging distributed systems, he strongly believes that distributed functional computing can significantly improves the productivty in building cloud scale distributed computing and storage systems, and is the way to go for building distributed platform in the future. A whitepaper on OneNet can be found at here.