On Mon, 1 Aug 2011, Kenneth Porter wrote:
--On Wednesday, July 20, 2011 12:54 AM -0300 Rogelio scubacuda@gmail.com wrote:
The free DHCP solution, ISC, seems to be having scaling issues (i.e. handling only about 200 DHCPDISCOVER and 20 DHCPRENEW requests)
I've only been watching this thread with half attention ---
DHCP is UDP based -- as such it does not handle collision retry logic on the server side (a server cannot resend what it does not know [due to a collision and drop] a client has requested of it)
So client retries are in play -- the ISC dhcpd server is able to handle loads substantially above the rates quoted per minute
I've had setups involving LTSP PXE booting of diskless workstations, supporting several hundred such clients all booting within 5 minutes of one another at a shift start
It is necessary to model, or perhaps view with tcpdump of ethereal, that traffic to see why the transfers are falling apart
-- Russ herrold