Microsoft’s AI infrastructure is expected to cut the cost of AI by reducing wasted effort.
What you need to keep in mind
- Microsoft’s AI infrastructure, codenamed Singularity, will reduce AI costs.
- Singularity enables hundreds of thousands of GPUs and AI accelerators to collaborate and minimize the quantity of time spent.
- Microsoft is making major investments in AI through $1 billion for OpenAI in the year 2019.
Microsoft is working to reduce the cost of artificial intelligence (AI) and the amount of time wasted in computing globally. A recent paper released in Microsoft’s Azure and Research teams examines Microsoft’s AI service, known as Singularity. This paper titled Singularity: Planet-Scale elastic and Preemptive scheduling of AI workloads (PDF) breaks its work for Microsoft at a technical level.
“Singularity is a fully-managed, distributed infrastructure service used to perform AI tasks within Microsoft that are compatible with various hardware accelerators. It was designed from scratch to scale across a range of hundreds of millions of GPUs and other AI accelerators,” explains Microsoft’s Azure and Research teams in their white paper. “Singularity was designed with one purpose in mind: reducing costs related to AI by maximizing the value of a predetermined amount of accelerators across the globe and also providing firm SLAs for a range of different pricing options. “
Microsoft’s Singularity lets the hundreds of thousands of GPUs with AI accelerators cooperate in simple English. Singularity is an infrastructure service designed to cut down on the amount of time and effort wasted. It views all the devices that are part of the infrastructure as a single cluster, ensuring that all devices are fully utilized.
Singularity can also prioritize various tasks. “While using the capacity, Singularity also provides some protection while keeping track of SLAs that are specific to jobs,” says Microsoft. “For example, Singularity adapts to increasing the inference task and frees capacity by scaling down dynamically or eliminating the training task from being performed. “
Contrary to other systems that require beginning from scratch following a failure, Singularity can return to the position where the task was ended. This drastically cuts down on wasted time as DNN tasks for training can take several weeks.
Microsoft has made substantial investments in AI over the years, culminating with the acquisition of one billion dollars into OpenAI in 2019. The Azure computer was ranked as one of the world’s top 10 most powerful supercomputers on November 20 20, 2021. Azure systems are used for large-scale computing, as well as machine learning.
As per ZDNet, Microsoft employed the Singularity codename to identify an unrelated initiative at one point. This Singularity operated as an operating system designed to run a microkernel.