RE: [dhcwg] dhc-server-mib

> -----Original Message-----
> From: Arvanitis Kostas
> Sent: Tuesday, May 27, 2003 05:22
>
*** Snip! ***
>
> 1. In the MIB, two notifications are defined, namely
> "dhcpv4ServerFreeAddressLow" and "dhcpv4ServerFreeAddressHigh", that deal
> with the free address count in shared nets. However, shared nets
> are stated
> as optional in the MIB, since some DHCP servers may not implement them.
> Shouldn't they be named something more specific, such as
> "dhcpv4SharedNetFreeAddressLow" and "dhcpv4SharedNetFreeAddressHigh"?
>
> 2. There are no notifications defined for subnets, although the required
> fields exist and are marked as "accessible-for-notify". Shouldn't
> there be
> "dhcpv4SubnetFreeAddressLow" and "dhcpv4SubnetFreeAddressHigh"
> definitions?
>

This design was based on practical experience with a widely deployed server,
in particular an implementation with tens of thousands of clients on
thousands of subnets.  Basically, trying to track the address counts to the
subnet level is both misleading and futile.  Most servers of which I'm aware
(and certainly the one on which this MIB design was based) do NOT allocate
leases by subnet, but rather by shared network

Suppose shared network A is composed of 3 subnets.  Using some opaque
algorithm, leases are assigned, and clients randomly return addresses,
eventually leaving a configuration that is largely non-deterministic based
on the starting conditions.  Shared network A may have 40% free leases,
although subnetwork 1 has only 3% free, while subnetwork 2 has 60% free, and
subnetwork 3 has 57% free.  It's important to only report counts that are
significant, and the impending unavailability of free leases on subnetwork 1
does NOT affect the ability of a typical server to hand out addresses for
shared network A, so I would argue that reporting counts by the subnet is
misleading, as it does not reflect the ability of the server to operate
correctly, and futile, because there is no control or action the
administrator can take to affect the count.

Hope that explanation helps.

--Barr

_______________________________________________
dhcwg mailing list
dhcwg@ietf.org
https://www1.ietf.org/mailman/listinfo/dhcwg