Re: Is Ganymede suitable for DHCP

["Jonathan Abbey" <jonabbey@arlut.utexas.edu>]

| I read with great interest your announcement of Ganymede.  The server
| is a very impressive piece of work.  As it happens, we are looking for
| a way to allow network administrators in various deparments to update
| their portions of a central DHCP database.  We were thinking of using
| Sybase with a web front-end, but I'm wondering now if Ganymede might
| be a better solution.
| 
| We would like to use static IP address assignments in DHCP, which
| means that we need to know the ethernet address of each machine, and
| enter that along with some other information into the database.
| Would it be possible to make ethernet address an internal data type
| in Ganymede, like IP address is now?  Ethernet addresses do require
| fairly complicated editing, as do IP addresses.

The reason that Ganymede doesn't use a special data type for
ethernet/mac addresses is that there is no need for being able to
break a mac address down hierarchically the way that you can with ip
addresses.

In the GASH schema we are using a simple string type for mac
addresses.. the server can fairly easily be customized to insure a
consistent form for ethernet addresses, after all.

We are also interested in melding DHCP with Ganymede for automated
provisioning of static IP addresses, but we haven't worked out all the
details yet.  I have good confidence that the server has what it takes
to be hooked up to a DHCP server, but for us at least, we'd likely have
difficulties trying to accurately track ethernet addresses to do
that.  We are working towards implementing something similar for NAT.

| We are using the DHCP server from ISC.  Its database is quite a bit
| more complex than DNS entries are.  There are a large number of options,
| and defaults are set in a hierarical manner.  Some of the complexity
| could be removed by excluding most of the options from the Ganymede
| database, although this also makes it less flexible.
| 
| I assume that Ganymede's permissions model would be okay.  We would
| like an individual network administrator to have update access only
| to DHCP entries for his own network, and then only for certain fields.
| For example, I would create the initial entries with empty ethernet
| address fields, and the network administrator would fill those in when
| he installed new PCs on his network.

Ganymede's default permission model would have no problem with that at
all.  In any case, if you are willing to write a java class or two,
the Ganymede server can enforce any sort of permissions logic you can
think of.

| I suppose that it would be a fair amount of work to configure Ganymede
| for this purpose.  Do you think it would be suitable?

The Ganymede system is easily flexible enough to handle everything
you're talking about.. the only questions would be whether you were
willing and able to write the java code required to parse your
existing data and to generate the data to be passed to the ISC server,
and whether the Ganymede server could efficiently handle the volume of
data you would be working with.  

At ARL we're handling data for a couple thousand systems, maybe two
score IP subnets, network connectivity information for about 600
rooms, plus complete NIS records for around 600 people.  I expect the
server today can comfortably handle 3 to 5 times that much data, but
no one has tried it yet, to my knowledge.

| -- 
| -Gary Mills-    -Unix Support-    -U of M Academic Computing and Networking-



----------------------------------------------------------------------------
To make changes to your subscription to the ganymede mailing list, send
mail to majordomo@arlut.utexas.edu.

To unsubcribe, include the line

unsubscribe ganymede

in the body of your mail message

Download ganymede from ftp.arlut.utexas.edu/pub/ganymede/.

To see the latest design specs for Ganymede, visit the Ganymede web page
at http://www.arlut.utexas.edu/gash2/
----------------------------------------------------------------------------