Skip to main content

Network

Introduction to networks used by private cloud virtual machines and physical machines.

Basic Concepts​

There are three basic network concepts, in logical order from bottom to top, they are VPC, layer 2 network and IP subnet.

VPC​

VPC (Virtual Private Cloud) refers to a virtual network isolation area within the cloud platform. Network traffic between VPCs is isolated. VPCs are further divided into classic networks and virtual networks.

Classic Network​

Classic network is also called flat network or Underlay. It refers to the physical network defined by physical switches, routers and network configurations within these physical devices. Virtual machines using classic networks rely on physical network devices to implement network communication between nodes. Communication between virtual machines under the same IP subnet relies on layer 2 switching. Communication between virtual machines under different IP subnets relies on layer 3 switching. In classic networks, layer 2 or layer 3 network switching between virtual machines located on different hosts, or layer 3 network switching between any virtual machines, all need to rely on physical network devices.

For unification, classic network is defined as a special default VPC (Default VPC). This VPC exists by default and represents the network isolation area corresponding to the physical network.

Virtual Network​

Virtual network is also called Overlay network. It is a virtual network composed of virtual network elements interconnected through tunnel technology (tunnel) built on physical networks. Virtual machines within virtual networks communicate through software-simulated virtual layer 2 switches and virtual routers (layer 3). You can create any number of virtual networks. Traffic between virtual networks is isolated.

Virtual networks are implemented based on open source ovn encapsulation.

Layer 2 Network (wire)​

In classic networks, a layer 2 network corresponds to a broadcast domain. Virtual machines in the same layer 2 network can directly communicate point-to-point without relying on layer 3 network device switching (but rely on layer 2 network devices, which may be virtual ovs bridges or physical switches). Physically, a layer 2 network usually corresponds to a network area under a layer 3 switch or router. There can be multiple layer 2 networks in classic networks.

In virtual networks, since any devices can directly communicate (through tunnels), it can be understood that the entire VPC is a layer 2 network. For model unification, the concept of layer 2 network is still retained in virtual networks, but wire and VPC can be considered equivalent. A layer 2 network will be created by default within a virtual network VPC.

IP Subnet (network)​

IP subnet corresponds to a subnet, which is a continuous IP address segment with the same subnet mask. IP subnets belong to layer 2 networks. IP subnets carry network configuration information, such as IP address range, subnet mask, gateway, VLAN ID, DNS, etc. If a virtual machine is connected to an IP subnet, the virtual machine's virtual network card must use an IP address allocated from the network segment of this IP subnet. The virtual network card's configuration is determined by this IP subnet's attributes.

In classic networks, each IP subnet can have its own VLAN ID. Virtual machines connected to this IP subnet must also use the corresponding VLAN ID.

Network Topology Constraints​

In classic networks, because communication between virtual machines relies on physical devices, which IP subnets virtual machines on specific hosts can access are constrained by the physical networks that hosts are connected to. That is, virtual machines on a host can only access IP subnets within the layer 2 networks that the host is connected to.

Virtual networks do not have network topology constraints. That is, any virtual machine on any host can access any VPC network.

📄️ Network Card Traffic Limit

This section introduces setting network card traffic limits. Network card traffic limits are divided into upstream and downstream traffic. As long as traffic on one side exceeds the limit, the network card will be cut off. Network card traffic limits are implemented by collecting network card receive and send packet monitoring data. The collection cycle is one minute, so traffic may exceed the limit but the network card will only be cut off after the traffic collection cycle arrives.