jw354 at cornell.edu
Fri Feb 13 15:54:52 UTC 2015
On Feb 9, 2015, at 7:57 AM, Vikas Jadhav wrote:
> I would like to know below things for dhcpd,
> *) Any alarm can generate after 80% pool utilization
> *) dhcpd application support upto how much TPS
> *) How max IP pool ranges can able to add
We scripted a rudimentary parser for the lease file
and compare it with the database we use to build
the pool configurations.
We tested with the dhcp perf tool to find out the TPS.
Re TPS performance, disk-write performance is key:
normal disk can handle a fair load, but really
fast disk write performance multiplies TPS
capacity. We settled on a smart controller with
NV RAM for the write cache and are very pleased.
Other choices are solid state disk, or RAM disk.
The purpose of the writes is to avoid losing address
assignments with server outages (i.e. keep things working),
a function that RAM disk undercuts, but you
can mitigate this by copying the data off the RAM
disk often, e.g. once per minute.
Making sure logging does not do a disk synch after
each transaction is also key to performance.
I believe the sum of all addresses in all the pools
directly affects the dhcpd initialization time.
It also affects the lease file size and the memory
More information about the dhcp-users