DHCP and True Redundancy

[Mr V]

Hi,

I am planning to migrate our DHCP server from a single NT 4 box to two
Windows 2003 servers.
I have been reading up on the subject and have looked at Q325473
http://support.microsoft.com/?id=325473
However, I still have questions in regard to redendancy that I can't seem to
find detailed anywhere.
I have read that Microsoft recommend an 80/20 balance of scope addresses
for two DHCP servers.
My question is this. How do I set two DHCP servers up, so when one falls
over, the other has a large enough pool of scope addresses to serve every
client on the network?
With an 80/20 ratio I would have to add another 0-254 range to cover all
clients if the 80% server died.
Basically how do I arrange true redundancy for DHCP that isn't simply 80/20?
Is it possible without creating an extra scope?

Thanks

Mr V

 

[Guest]

There are afew ways of having redundancy for DHCP, namely;
1. clustering of DHCP
2. using the 80/20 rule
3. having a standby DHCP server

I don't believe any organization is willing to pay for the price of
clustering just for DHCP, simply not justifiable.

Using the 80/20 rule is the usual normal when you want 2 servers in
different subnets to support each other. A lot of people wonder how to
support all the clients with just 20% of the IP address if the primary server
fails. Remember that in the organization, we basically using private
addresses range reserved by the IANA. In this case you should use either
class A or class B to assign addresses to clients. Using class C is never a
good planning bevause it can't scale. With 16 mil and 4 mil hosts available
respectively, it is more than enough for you to plan your 80/20 rule yet able
to cover for all clients.

Having a standby server as DHCP can be shared with existing server (file or
print or etc) without having to create additional scopes or incur extra
costs. The only issue to take note is when the primary server fail and the
standby server is brought up, the new server when lease address to new
clients might experience ip conflict where the address is being used by
existing clients leased out by the fail server. To resolve this, either make
use of the ip address conflict detection feature in the DHCP or order all
clients to reboot their machines so to go through the 4 discovery phases
again.

HTH.

 

[Mr V]

There are afew ways of having redundancy for DHCP, namely;
1. clustering of DHCP
2. using the 80/20 rule
3. having a standby DHCP server

I don't believe any organization is willing to pay for the price of
clustering just for DHCP, simply not justifiable.

Using the 80/20 rule is the usual normal when you want 2 servers in
different subnets to support each other. A lot of people wonder how to
support all the clients with just 20% of the IP address if the primary
server
fails. Remember that in the organization, we basically using private
addresses range reserved by the IANA. In this case you should use either
class A or class B to assign addresses to clients. Using class C is never
a
good planning bevause it can't scale. With 16 mil and 4 mil hosts
available
respectively, it is more than enough for you to plan your 80/20 rule yet
able
to cover for all clients.

Having a standby server as DHCP can be shared with existing server (file
or
print or etc) without having to create additional scopes or incur extra
costs. The only issue to take note is when the primary server fail and the
standby server is brought up, the new server when lease address to new
clients might experience ip conflict where the address is being used by
existing clients leased out by the fail server. To resolve this, either
make
use of the ip address conflict detection feature in the DHCP or order all
clients to reboot their machines so to go through the 4 discovery phases
again.

HTH.

Thanks for your reply. Your comments have bee noted.

Mr V