This is a new feature, which came with NetScaler 10.1. It is the ability to load balance TFTP servers. Before the 10.1 release of NetScaler, this required a great deal of work including the use of Direct Server Return (DSR) and other options. However, they are no longer required.

An important point to remember is that when you boot a virtual machine using PVS, it uses either PXE or the DHCP options, including options 66 and 67. This guide uses DHCP options to distribute the link to the bootstrap file.

Now, in order to set up load balancing for TFTP properly, we need a monitor that we can use to verify if load balancing is operational. To get a monitor for TFTP, we can follow the guidelines located at http://blogs.citrix.com/2011/01/06/monitoring-tftp-with-citrix-NetScaler-revisited-pvs-edition/.

After we have created the monitor, we need to add the servers where the TFTP service resides. Next, we need to create a service for each TFTP server. Here, we need to choose TFTP as the protocol, and in the Port field we need to enter 69. Then, we need to bind the custom-made monitor to the service. After this is done, we create the load-balanced vServer, where we enter an IP address, the name, the protocol (TFTP), and the port (69). When this is complete and we have created the vServer, we can alter the DHCP option 66 to point to the new VIP address that we created in NetScaler.

As a side note, it is also possible to deliver the bootstrap using HTTP instead of TFTP. This scenario is only viable for XenServer as it uses gPXE, which allows for extra features such as HTTP. This makes it a lot easier to load balance, as we only need to load balance a simple HTTP server, and change the option 67 boot filename to point to http://serverip/ARDBP32.BIN. However, this is not supported by Citrix and should only be used in environments where HTTP is a more suited protocol. As always, remember to save the configuration using the GUI or the save config command in CLI.