If you have a number of routers/sites that have direct connectivity between themselves and they are geographically closer to each other than back to your hub, one option you have is to set up a router as the TFTP server. If you can reduce the latency, then you can transfer more blocks in any given period of time. If you have a round-trip latency of 100ms between your client and server then you can at most transfer 10-blocks of data per second. Slow tranfer over the WAN is sometimes caused by low bandwidth, but more usually it's due to high latency. There are measurements in the aforementioned RFC that show a 2.8x performance improvement when using 1428-byte blocks compared to 512-byte blocks.

The default TFTP blocksize is 512-bytes, but if your client and server both support RFC 2348 TFTP Blocksize Option then the blocksize can be increased to approximately 1450-bytes.

With this in mind there are really only two ways to increase TFTP performance. As you may be aware, the way TFTP works is that it transfers a single data packet containing one block of data, and that packet must be acknowledged before the next packet can be sent.

Download tftpd64 cisco