Basic Network selection when using ESX and vmotion

This article is from the network, not original CC

VMware ESX Server and VMWare Infrastructure 3 provide network and system administrators with multiple options to create network configurations.

However, flexibility always brings complexity. This article describes the recommended basic network options when using ESX and vmotion.

Network Configuration of ESX host

The minimum requirement for installing ESX on a server is to have a separate network adapter, but in a standard data center environment, at least two NICs (network adapter or network adapter) are required) is the minimum configuration requirement for redundancy. In other aspects, network adapters are also required by the ESX console operating system to communicate with external networks. Additionally, you may need to install additional network adapters for other options and configurations, depending on the server hardware used (rack or blade servers are discussed below ).

Tower and rack servers

When a tower or rack server is used, the best configuration requires at least five network adapters. To properly separate network traffic, three adapters are required:

Management Network of the ESX console operating system (one Nic)

Vmkernel Network: vmotion and (or) iSCSI and NFS are required.

Vmotion requires at least one network interface with a speed of 1 GB per second. The best solution is to put it in a separate network to minimize the latency and network congestion of vmmotion (a NIC)

If iSCSI is used, a virtual Nic must be created for the kernel to use iscisi (the best solution is to use two NICs for redundancy ). Iscsi should be physically separated to ensure bandwidth, reduce contention, and ensure security.

NFS can be used to access the VM documentation library. If NFS is used, another network port must be used and allocated to NFS.

Virtual Machine Network Interface (at least two NICs are used for redundancy)

Blade Server

The blade server is usually configured with the integrated high-order integrated switch (chassis switch), which limits the number of network adapters that each blade connects to the high-order integrated switch. Depending on the blade system vendor, this limit may reach eight NICs or HbA (host bus adapter) (early blade servers are typically limited to two or four ).

Similarly, in blade servers and high-end integrated switches, the number of uplink ports from high-end integrated switches to distributed/Core switches is limited. The network administrator should configure the uplink ports to be aggregated from the high-end integrated switches and implement 802.1Q VLAN tags. It is recommended that the network should be at least 1 GB per second. This will allow ESX to provide logical separation of different types of network traffic and have sufficient bandwidth to run all the operations effectively.

If you use ESX in a blade server environment, we recommend that you have at least 1 GB of network per second. In this way, the network bandwidth competition from the host to the enterprise's sub-operating system will be minimized. For high-speed storage requirements, I suggest using an optical fiber storage network solution instead of iSCSI (unless 10 Gb iSCSI is used per second ), this is because the bandwidth of data from all hosts and sub-operating systems is limited to that of the high-end integrated switches, and in this shared environment, I/O may become a bottleneck. Running an iSCSI storage infrastructure will reduce the available network bandwidth for data connections and may affect the performance of all systems of the high-end integrated switches.

Esx nic Management

The ESX host can be configured to use multiple Ethernet ports in multiple ways. The port can be set to active/standby. If a physical port is disconnected, ESX detects the port and uses the second configured port. Multiple configured ports can be set as standby. In this way, the administrator can use a single uplink connection for non-critical components, such as the console operating system. However, a port is retained to the server for backup to prevent major port faults.

In addition, the port can be configured with multiple Server Load balancer configurations (Server Load balancer is only used for outbound traffic. The three Server Load balancer configurations are based on the source port ID, the hash value of the source MAC address, and the hash value of the source IP address and the destination IP address (see the detailed explanation in the VMware documentation ). You can detect a fault by monitoring the connection status of the adapter and finding a fault in the upstream network using a signal. A port can also be used to notify the vswitch that a port has been reconfigured. Therefore, the ARP table is updated to minimize network errors.

With these options, ESX provides the flexibility to optimize the configuration and virtual network architecture to meet many different needs. Depending on your needs and hardware constraints, these options allow you to configure the ESX environment within the budget to provide the most robust and practical solutions.