Feature comparison
SkyPilot is a dynamic framework designed to run AI and batch workloads on virtually any infrastructure. It offers unified execution, substantial cost savings, and unmatched GPU availability. By abstracting away the complexities of infrastructure management, SkyPilot lets you effortlessly launch development clusters, manage jobs, and serve applications on any platform.
With SkyPilot, you can efficiently queue, run, and auto-recover multiple jobs. It supports a wide array of clusters, clouds, and hardware - including reserved GPUs, Kubernetes clusters, and over 12 different cloud providers. Its flexible provisioning system handles GPUs, TPUs, and CPUs, all with built-in auto-retry mechanisms to keep your operations running smoothly.
SkyPilot optimizes your cloud usage by automatically shutting down idle resources with its Autostop feature. It achieves 3-6x cost savings through Managed Spot instances that recover from preemptions and can double your savings by automatically selecting the most cost-effective and available infrastructure. Best of all, SkyPilot fully supports your existing GPU, TPU, and CPU workloads without requiring any code changes. It integrates seamlessly with popular platforms like Kubernetes, AWS, GCP, Azure, and more.
Features | ![]() | ![]() |
---|---|---|
Supported WorkloadsWhat kind of workloads can you run on these platforms | Cloudfleet covers a broad spectrum of workloads, including serverless microservices, web applications, and batch processing. It also supports various hardware accelerators, giving you the flexibility to run diverse applications. | Primarily focuses on AI and batch workloads, supporting GPUs, TPUs, and CPUs across multiple clouds and hardware platforms. |
Multi-Cloud SupportCan you provision nodes across multiple cloud providers | Both Cloudfleet and SkyPilot support over 12 cloud providers, giving you the flexibility to choose the best infrastructure for your needs. | |
Multi-Region SupportIs it possible to deploy applications across multiple regions | Yes, even within a single cloud provider, Cloudfleet can deploy nodes in multiple regions, enhancing availability and performance. | Absolutely. SkyPilot can deploy tasks and even instances of a single task across multiple regions, ensuring optimal resource utilization. |
Hybrid SupportCan you deploy nodes both on-premises and in the cloud | Designed with hybrid environments in mind, all Cloudfleet clusters can accept nodes from any cloud provider or on-premises setup. | Deploys a lightweight Kubernetes cluster on remote machines using k3s, offering some level of hybrid support. |
Managed Spot InstancesHow do they handle spot and on-demand instances based on availability | When Spot or Preemptible instances are terminated, Cloudfleet automatically shifts workloads to on-demand or other types of spot instances, ensuring uninterrupted service. | Currently, in model serving mode, SkyPilot relies on the user to retry tasks in the event of spot instance preemptions. |
Identity and Access ManagementWhat about user management and security controls | Offers a built-in identity and access management system. You can manage users and permissions, configure identity federation, and set up two-factor authentication. | Does not provide built-in IAM features. Operators need to set up their own identity and access management systems. |
Managed ServiceHow easy is it to deploy and manage these platforms | A fully managed SaaS platform that requires no manual configuration. It’s designed for ease of use so you can focus on your workloads. | A Python package that you need to deploy and manage manually, which might require more hands-on effort. |
Industry Standard APIDo they offer access via common APIs | Utilizes the standard Kubernetes API, making it compatible with any Kubernetes client or tool you might already be using. | Uses a custom YAML interface to specify tasks, including resource requirements, setup commands, and file mounts. |
Managed Cloud AccountsCan you access cloud providers without opening individual accounts | Cloudfleet’s Connected Fleets feature allows you to use cloud infrastructure from all major providers without the hassle of opening individual accounts, handling quotas, or managing support requests. | Requires you to open accounts with each supported cloud provider to use the service, which can be time-consuming and complex. |
Cloudfleet vs SkyPilot
Cloudfleet Kubernetes Engine (CFKE) is a next-generation managed Kubernetes service that offers a simplified and cost-effective solution for deploying, managing, and scaling containerized applications. Designed to streamline Kubernetes cluster deployment and management, CFKE provides a seamless and efficient experience for software teams. By leveraging CFKE, you can focus on developing and optimizing your applications while it handles the complexities of cluster management and infrastructure provisioning.
By utilizing a state-of-the-art technology stack, CFKE enables unique use cases and removes the need to operate the cluster or manage cluster federation.
Whether you’re looking to optimize costs, simplify cluster management, or build new products, CFKE fits perfectly into your existing stack.
User management and access control are built-in, and you can easily integrate with your existing identity provider. CFKE also provides comprehensive monitoring and logging tools, giving you insights into your applications’ performance and health.
CFKE handles the provisioning and scaling of the Kubernetes control plane - including API servers and the backend persistence layer - across different cloud providers, regions, and availability zones. This ensures high availability and fault tolerance. It automatically detects and replaces unhealthy control plane nodes, applies patches, and provides serverless compute for containers, eliminating the need to manage servers manually. You pay only for the resources your applications use, while also benefiting from enhanced security through application isolation by design.
Getting started with Cloudfleet is easy, and you can begin with our free tier that covers control plane deployments. It allows you to use cloud infrastructure from both Cloudfleet and customer-managed cloud provider accounts, as well as self-managed on-premises nodes.
Comprehensive support for major cloud providers and regions, ready for multi-cloud deployments.
Expand servers, increase storage, or migrate to a different provider with zero downtime - at the push of a button.
High availability is included in the open-source feature set, ensuring reliability for mission-critical applications.
Monitor and manage your cloud infrastructure effortlessly with your preferred tools.
Expert 24/7 support, available 365 days a year. We're here whenever you need us.
End-to-end encryption, isolated networking, dedicated VMs, and full compliance certifications.
Create your free Cloudfleet Kubernetes cluster in minutes - no setup hassle, no cost. Get started instantly with the always-free Basic plan.
Get started →