Which Scheduler to Use? Slurm vs Kubernetes

Kubernetes is more powerful than Slurm for containerized workloads, but it doesn’t have the built-in scheduling capabilities that AI/ML requires. That’s why Run:AI built a scheduler specifically for AI.

Slurm & Kubernetes Not Built For AI

Slurm is the go-to scheduler for managing the distributed, batch-oriented workloads typical of HPC. Kubernetes is the go-to for the management of flexible, containerized workloads and microservices. 

These requirements are essential, but they’re not the only demands when it comes to AI/ML workloads. Both schedulers lack critical high-performance scheduling components like batch scheduling, preemption, and multiple queues for efficiently orchestrating long running jobs.

AI Deserves A Purpose-Built Scheduler

Bridging the efficiency of High-Performance Computing and the simplicity of Kubernetes, the Run:AI Scheduler allows users to easily make use of fractional GPUs, integer GPUs, and multiple-nodes of GPUs, for distributed training on Kubernetes.

Built as a plug-in to K8s with all of the scheduling capabilities specific to deep learning, Run:AI requires no advanced setup, and works with any number of Kubernetes orchestration “flavors” such as Vanilla, RedHat OpenShift and HPE Container Platform.

The AI Infrastructure Stack of the Future

The Run:AI Stack
From fractional GPU to multi-node distributed computing
Kubernetes-based scheduler
Seamless for data scientists
Control and visibility

Gain Visibility into GPU Consumption

With Run:AI’s flexible ‘virtual pool’ of compute resources, IT can visualize their full infrastructure capacity and utilization across sites, whether on prem or in the cloud. The Run:AI GUI offers a holistic view of GPU infrastructure utilization, usage patterns, workload wait times, and costs.

Take Control of Training Times and Costs

Run:AI’s cloud-native scheduling mechanism enables IT to easily define and set policies for consumption of GPU compute in line with business goals. IT gains full control over GPU utilization, with advanced monitoring tools, queueing mechanisms, and automatic preemption of jobs.

Run Data Experiments at Maximum Speed

Run:AI provides data scientists with optimal speeds for training and inference. By abstracting AI workloads from compute power, while honoring a guaranteed quota of GPUs for each project, enterprises see faster results from DL modeling.

Optimize Deep Learning Training

Run:AI optimizes utilization of your existing infrastructure by distributing workloads in an ‘elastic’ way – dynamically changing the number of resources allocated to a job – allowing data science teams to run more experiments on the same hardware.

Learn the essential capabilities to consider when choosing the best AI scheduling solution

Trusted by IT and loved by Data Scientists at:

Rapid AI development is what this is all about for us. What Run:AI helps us do is to move from a company doing pure research, to a company with results in production.

Siddharth Sharma, Sr. Research Engineer, Wayve

With Run:AI we’ve seen great improvements in speed of experimentation and GPU hardware utilization. Reducing time to results ensures we can ask and answer more critical questions about people’s health and lives.

M. Jorge Cardoso, Associate Professor & Senior Lecturer in AI at King’s College London


Run on-premises or in the cloud

Policy-based automated scheduling

Optimize utilization of costly resources

Elastic virtual pools of GPU

Full control, visibility and prioritization

No code changes required by the user

1-click execution of experiments

Kubernetes plug-in simple integration

The Essential Guide to GPU Machine Scheduling

Discover the challenges of working with GPUs for AI and the solutions emerging from the worlds of virtualization, HPC and distributed computing.

© Run:AI 2021 | info@run.ai

We use cookies on our site to give you the best experience possible. By continuing to browse the site, you agree to this use. For more information on how we use cookies, see our Privacy Policy.