[syslinux] tftp-hpa performance?

[Dyks, Axel (XL)]

H. Peter Anvin wrote:
> Dyks, Axel (XL) wrote:
>>> Why?
>> Which part is "why?" refering to?
>>
>> * standalone is prefered: less overhead
>>   // ... assuming that in the poster's case (xinetd's) security/tcp-wrapper features
>>   // are less important than having maximum performance / mininum latency)
>>
>> * distro's default config: It's default on gentoo, don't know much about others
>>
> 
> Given how tftp-hpa works, the extra overhead is negible, and in fact
> approaches zero for a busy enough server.

Well, I think it would be interesting to figure out in a large and "busy"
environment, whether there is a difference or not.

Having in mind that someone posted here that launching "tftpd" from "inetd"
didn't work while standalone mode did (I think, this happened on Suse 9.X),
it might be a good idea to get things running in "standalone mode" and try
"[x]inetd mode" later.

Actually I have access to a "perfect test environment" (a customer with about
5000 clients), but I doubt that they will let me "touch" their proprietary
"intel-pxe-tftp" server and use "tftp-hpa" instead.
// Though hpa-tftp is working fine in the test environment and
// pxelinux+kernel+initrd is already used in production
// to install (imaging+sysprep) all Windows XP clients.

:-) Axel