Highlights –

  • Singularity may be the next phase in turning Brainwave into a commercially available service by Microsoft.
  • The new AI infrastructure service provides data scientists and AI practitioners with a way to build, scale, experiment, and iterate their module on a Microsoft-provided distributed infrastructure service explicitly built for AI.

Microsoft’s Azure and Research teams are working to build a new AI infrastructure service, codenamed Singularity. The Singularity team is building a solution described as “a new AI platform service ground-up from scratch that will become a major driver for AI, both inside Microsoft and outside,” by Microsoft.

The project team published a paper titled, ‘Singularity: Planet-Scale, Preemptible and Elastic Scheduling of AI Workloads’ that elaborates on technical information and details of the new service. The Singularity service was designed to help data scientists and AI practitioners with a way to build, scale, experiment, and iterate their models on a Microsoft-provided distributed infrastructure service explicitly built for AI.

Mark Russinovich, Azure Chief Technical Officer; Partner Architect Rimma Nehme, who has also worked on Azure Cosmos DB before joining Azure to work on AI and deep learning in 2019; and Technical Fellow Dharma Shukla are authors of the newly published paper.

Excerpts from the paper – “At the heart of Singularity is a novel, workload-aware scheduler that can transparently preempt and elastically scale deep learning workloads to drive high utilization without impacting their correctness or performance, across a global fleet of accelerators (e.g., GPUs, FPGAs).”

Earlier, Microsoft had also made plans to provide Field-Programmable Gate Arrays (FPGAs), or field-programmable gate arrays, to customers as a service. In 2018, it had also revealed its ‘Project Brainwave’ work, designed to provide fast AI processing in Azure. The tech giant also disclosed a preview of Brainwave powered Azure Machine Learning Hardware Accelerated Models in the cloud. This marked the first step in making FPGA processing available to customers for AI workloads.

Singularity may be the next phase in turning Brainwave into a commercially available service by Microsoft.

In 2019, Microsoft had invested USD one billion in OpenAI, and a year later, officials announced that they had built the fifth most powerful publicly recorded supercomputer in partnership with and exclusively for OpenAI.

While Microsoft’s AI supercomputer is built exclusively for OpenAI, officials have stated that they intend to make the company’s huge AI models and training optimization tools public through Azure AI services and GitHub.

Under its banner ‘Azure AI,’ Microsoft has also made various accelerators and services available for those customers who do not require a dedicated supercomputer. In November 2021, Microsoft expanded its AI supercomputer lineup with 80GB NVIDIA A100 GPUs in Azure.

To recall, Microsoft used the Singularity codename for another Microsoft Research project earlier. The earlier one was a microkernel OS and a set of related tools and libraries designed entirely in managed code.

The Singularity project was built from scratch as a proof-of-concept and in no way based on Windows. The project ended up spawning and/or influencing several other OS research projects at Microsoft, including Helios, Midori, Barrelfish, and Drawbridge.

Moreover, Microsoft is not the only tech giant trying to make AI supercomputing capabilities available internally and to customers. Meta (formerly known as Facebook) is also working on it by positioning its work as the key to unlocking the metaverse.