English | 简体中文
We are a Cloud Native Computing Foundation sandbox project
Spiderpool is the underlay and RDMA network solution of the Kubernetes, for bare metal, VM and any public cloud
Spiderpool is an underlay and RDMA network solution for the Kubernetes. It enhances the capabilities of Macvlan CNI, ipvlan CNI, SR-IOV CNI, fulfills various networking needs, and supports to run on bare metal, virtual machine, and public cloud environments. Spiderpool delivers exceptional network performance, particularly benefiting network I/O-intensive and low-latency applications like storage, middleware, and AI. It could refer to website for more details.
The underlay CNI is mainly including macvlan, ipvlan, and SR-IOV, which cloud access the layer 2 network of the node. It has some advantages:
-
macvlan, ipvlan, and SR-IOV is crucial for supporting RDMA network acceleration. RDMA significantly enhances performance for AI applicaitons, latency-sensitive and network I/O-intensive applications, surpassing overlay network solutions in terms of network performance.
-
Unlike CNI solutions based on veth virtual interfaces, underlay networks eliminate layer 3 network forwarding on the host, avoiding tunnel encapsulation overhead. This translates to excellent network performance with high throughput, low latency, and reduced CPU utilization for network forwarding.
-
Connecting seamlessly with underlay layer 2 VLAN networks enables both layer 2 and layer 3 communication for applications. It supports multicast and broadcast communication, while allowing packets to be controlled by firewalls.
-
Data packages carry the actual IP addresses of Pods, enabling direct north-south communication based on Pod IPs. This connectivity across multi-cloud networks enhances flexibility and ease of use.
-
Underlay CNI can create virtual interfaces using different parent network interfaces on the host, providing isolated subnets for applications with high network overhead, such as storage and observability.
-
Simplified installation and usage
Through eliminating the need for manually installing multiple components such as Multus CNI, RDMA and SR-IOV, Spiderpool simplifies the installation process and decreases the number of running PODs. It provides streamlined installation procedures, encapsulates relevant CRDs, and offers comprehensive documentation for easy setup and management.
-
CRD-based dual-stack IPAM
Spiderpool provides exclusive and shared IP address pools, supporting various affinity settings. It supports to assign static IP addresses for stateful applications such as mysql, redis, and kubevirt, while enabling fixed IP address ranges for stateless ones. Spiderpool automates the management of exclusive IP pools, ensuring excellent IP reclamation to avoid IP leakage. In additions, it provides wonderful IPAM performance.
The IPAM of Spiderpool could be available for any main CNI supporting third-party IPAM plugin, not only including Macvlan CNI, ipvlan CNI, and SR-IOV CNI, but also calico and weave as static IP usage.
-
Multiple network interfaces from underlay and overlay CNI
Spiderpool enables scenarios where Pods can have multiple underlay CNI interfaces or a combination of overlay and underlay CNI interfaces. It ensures proper IP addressing for each CNI interface and effectively manages policy routing to maintain consistent data paths, eliminating packet loss concerns. It could strengthen cilium, calico, and kubevirt.
-
Enhanced network connectivity
As we know, native CNI of macvlan ipvlan SR-IOV has lots of communication limits. However, Spiderpool establishes seamless connectivity between Pods and host machines, ensuring smooth functioning of Pod health checks. It enables Pods to access services through kube-proxy or eBPF-based kube-proxy replacement. Additionally, it supports advanced features like IP conflict detection and gateway reachability checks. The network of Multi-cluster could be connected by a same underlay network, or Submariner.
-
eBPF enhancements
The eBPF-based kube-proxy replacement significantly accelerates service access, while socket short-circuiting technology improves local Pod communication efficiency within the same node. Compared with kube-proxy manner, the improvement of the performance is Up to 25% on network delay, up to 50% on network throughput.
-
RDMA support
Spiderpool offers RDMA solutions based on RoCE and InfiniBand technologies. POD could use the RDMA device in shared or exclusive mode, and it is fit for AI workloads.
-
Dual-stack network support
Spiderpool supports IPv4-only, IPv6-only, and dual-stack environments.
-
Good network performance of latency and throughput
Spiderpool performs better than overlay CNI on network latency and throughput, refer to performance report.
-
Metrics
Spiderpool, powered by underlay CNI, offers unparalleled network performance compared to overlay CNI solutions, as evidenced in I/O Performance. It can be effectively applied in various scenarios, including:
-
Support to deploy on the environments of bare metal, virtual machine, and public cloud, especially offer a unified underlay CNI solution for hybrid cloud.
-
Traditional host applications. They hope to directly use the underlay network, with the reasons such as direct access to the underlay multi-subnet, multicast, multicast, layer 2 network communication, etc. They cannot accept the NAT of the overlay network and hope to seamlessly migrate to Kubernetes.
-
Network I/O-intensive applications such as middleware, data storage, log observability, and AI training.
-
Applications which needs a separate network bandwidth.
-
Latency-sensitive application.
-
Refer to Quick start to explore Spiderpool quickly.
-
Refer to Usage Index for usage details.
-
Refer to Spiderpool Architecture for more detailed information.
| Features | macvlan | ipvlan | SR-IOV |
|---|---|---|---|
| Service By Kubeproxy | Beta | Beta | Beta |
| Service By Kubeproxy Replacement | Alpha | Alpha | Alpha |
| Network Policy | In-plan | Alpha | In-plan |
| Bandwidth | In-plan | Alpha | In-plan |
| RDMA | Alpha | Alpha | Alpha |
| IPAM | Beta | Beta | Beta |
| Multi-Cluster | Alpha | Alpha | Alpha |
| Egress Policy | Alpha | Alpha | Alpha |
| Multiple NIC And Routing Coordination | Beta | Beta | Beta |
| Scenarios | Bare metal | Bare metal and VM | Bare metal |
For detailed information about all the planned features, please refer to the roadmap.
Refer to Blogs.
The project is governed by a group of Maintainers and Committers. How they are selected and govern is outlined in our Governance Document.
A list of adopters who are deploying Spiderpool in production, and of their use cases, can be found in file.
Refer to Contribution to join us for developing Spiderppol.
The Spiderpool community is committed to fostering an open and welcoming environment, with several ways to engage with other users and developers. You can find out more information by visiting our community repository.
-
Slack: join the #Spiderpool channel on CNCF Slack by requesting an invitation from CNCF Slack. Once you have access to CNCF Slack, you can join the Spiderpool channel.
-
Community Meeting: Welcome to our community meeting held on the 1st of every month. Feel free to join and discuss any questions or topics related to Spiderpool.
-
Email: refer to the MAINTAINERS.md to find the email addresses of all maintainers. Feel free to contact them via email to report any issues or ask questions.
-
WeChat group: scan the QR code below to join the Spiderpool technical discussion group and engage in further conversations with us.
Spiderpool is licensed under the Apache License, Version 2.0. See LICENSE for the full license text.
Copyright The Spiderpool Authors
We are a Cloud Native Computing Foundation sandbox project.
The Linux Foundation® (TLF) has registered trademarks and uses trademarks. For a list of TLF trademarks, see Trademark Usage.
