Network Address Translation (NAT) gateways can be divided into the following types: Small, Middle, Large, and Super Large-1. The type of NAT gateway that you choose determines the maximum number of Source Network Address Translation (SNAT) connections and the number of new SNAT connections per second. However, it does not affect the performance of Destination Network Address Translation (DNAT).
The following table lists the types of NAT gateway.
|Type||Maximum number of SNAT connections||Number of new SNAT connections per second|
- The bandwidth and the number of IP addresses in a NAT service plan are not restricted by the type of NAT gateway that you choose.
- CloudMonitor monitors only the maximum number of SNAT connections for NAT gateways. It does not monitor the number of new SNAT connections per second.
- The timeout of SNAT connections in a NAT gateway is 900 seconds.
- To avoid the timeout of SNAT connections caused by network congestion and Internet instability, make sure that your applications support automatic reconnection, which ensures higher availability.
- NAT gateways do not support packet fragmentation.
- For the same destination public IP address and port, the number of Elastic IP addresses configured for a NAT gateway determines the maximum number of concurrent connections. If an individual Elastic IP address is bound to the NAT gateway, the maximum number of connections is 55,000. If N Elastic IP addresses are bound to the NAT gateway, the maximum number of connections is increased to N × 55,000.
- Assume that you have multiple ECS instances deployed in a VPC network and the ECS instances are not assigned public IP addresses. The ECS instances access the same destination IP address and port on the Internet through a NAT gateway at a bandwidth higher than 2 Gbit/s. To avoid packet loss caused by the upper limit of ports for a single public IP address, we recommend that you bind 4 to 8 public IP addresses to the NAT gateway and create a SNAT pool.