Introduction
Cloudland is a lightweight Infrastructure-as-a-Service (IaaS) framework designed to manage virtual machine instances, software-defined networks (SDN), storage volumes, and more. Capable of supporting over 10,000 hypervisors in a single cluster, Cloudland is ideal for large-scale public cloud environments.
The goal of Cloudland is to provide a simple, extensible IaaS platform with easy-to-use operations that save time for both developers and operators.
Cloudland’s core functionality revolves around provisioning virtual machine instances using KVM. This provides a flexible base for deploying Platform-as-a-Service (PaaS) solutions, such as Kubernetes, on top of the infrastructure.
Cloudland includes a built-in SDN solution based on VXLAN, offering scalable, isolated network segments. It also features a distributed virtual router for Network Address Translation (NAT), ensuring efficient and secure traffic routing across virtual networks.
Cloudland uses security groups based on iptables to provide flexible and granular control over network traffic, ensuring secure communication between virtual machines and the internet
The architecture is mainly based on HPC (High Performance Computing) technology SCI (Scalable Communication Infrastructure) which is an Eclipse OpenSource project https://wiki.eclipse.org/PTP/designs/SCI.
Cloudland has demonstrated exceptional performance in large-scale environments. Benchmarks show an average of ~50 VMs per hypervisor per minute during instance launches.
In long-term tests, Cloudland successfully handled 500k VMs across 20,000 hypervisors, with a success rate of over 99.99%.
Cloudland's SCI based architecture ensures high availability and reliability through automatic failover and recovery. Its frontend, agents, and backends use heartbeat checks to monitor system health in real time. In case of failure, Cloudland can auto-reconnect and recover, ensuring minimal disruption.
Cloudland stands out for its simplicity. Unlike many IaaS solutions, it doesn’t rely on a complex array of components. The straightforward development approach results in a flat learning curve for developers, while its simple deployment and stable operation make it easy to maintain with minimal effort from operators.