5. Operational and Management Considerations
This section covers two particular areas of operations and
management: configuration requirements and transition from or
coexistence with the MIB module in [RFC4008].
5.1. Configuration Requirements
This MIB module assumes that the following information is configured
on the NAT device by means outside the scope of the present document
or is imposed by the implementation:
o the set of address realms to which the device connects;
o for the CGN application, per-subscriber information including
subscriber index, address realm, assigned prefix or address, and
(possibly) policies regarding address pool selection in the
various possible address realms to which the subscriber may
connect. In the particular case of DS-Lite [RFC6333] access, as
well as the assigned outer-layer (IPv6) prefix or address, the
subscriber information will include an inner (IPv4) source
address, usually 192.0.0.2;
o the set of NAT instances running on the device, identified by NAT
instance index and name;
o the port mapping, filtering, pooling, and fragment behavior for
each NAT instance;
o the set of protocols supported by each NAT instance;
o for the pooled NAT and CGN applications, address pool information
for each NAT instance, including for each pool the pool index,
address realm, address type, minimum and maximum port number, the
address ranges assigned to that pool, and policies for access to
that pool's resources;
o static address and port map entries.
As described in previous sections, this MIB module does provide read-
write objects for control of notifications (see especially
Section 3.1.2) and limiting of resource consumption (Section 3.1.1).
This document is written in advance of any practical experience with
the setting of these values and can thus provide only general
principles for how to set them.
By default, the MIB module definition disables notifications until
they are explicitly enabled by the operator, using the associated
threshold value to do so. To make use of the notifications, the
operator may wish to take the following considerations into account.
Except for the low address pool utilization notification, the
notifications imply that some sort of administrative action is
required to mitigate an impending shortage of a particular resource.
The choice of value for the triggering threshold needs to take two
factors into account: the volatility of usage of the given resource,
and the amount of time the operator needs to mitigate the potential
overload situation. That time could vary from almost immediate to
several weeks required to order and install new hardware or software.
To give a numeric example, if average utilization is going up 1% per
week but can vary 10% around that average in any given hour, and it
takes two weeks to carry through mitigating measures, the threshold
should be set to 88% of the corresponding limit (two weeks' growth
plus 10% volatility margin). If mitigating measures can be carried
out immediately, this can rise to 90%. For this particular example,
that change is insignificant, but in other cases the difference may
be large enough to matter in terms of reduced load on the management
The notification rate-limit settings really depend on the operator's
processes but are a tradeoff between reliably reporting the notified
condition and not having it overload the management plane.
Reliability rises in importance with the importance of the resource
involved. Thus, the default notification intervals defined in this
MIB module range from 10 seconds (high reliability) for the address
and port map entry thresholds up to 60 seconds (lower reliability)
for the per-subscriber port entry thresholds. Experience may suggest
The limits on number of instance-level address map and port map
entries and held fragments relate directly to memory allocations for
these tables. The relationship between number of map entries or
number of held fragments and memory required will be implementation-
specific. Hence it is up to the implementor to provide specific
advice on the setting of these limits.
The limit on simultaneous number of active subscribers is indirectly
related to memory consumption for map entries, but also to processor
usage by the NAT instance. The best strategy for setting this limit
would seem to be to leave it disabled during an initial period while
observing device processor utilization, then to implement a trial
setting while observing the number of blocked packets affected by the
new limit. The setting may vary by NAT instance if a suitable
estimator of likely load (e.g., total number of hosts served by that
instance) is available.
5.2. Transition from and Coexistence with NAT-MIB (RFC 4008)
A manager may have to deal with a mixture of devices supporting the
NAT-MIB module [RFC4008] and the NATV2-MIB module defined in the
present document. It is even possible that both modules are
supported on the same device. The following discussion brings out
the limits of comparability between the two MIB modules. A first
point to note is that NAT-MIB is primarily focused on configuration,
while NATV2-MIB is primarily focused on measurements.
To summarize the model used by [RFC4008]:
o The basic unit of NAT configuration is the interface.
o An interface connects to a single realm, either "private" or
"public". In principle that means there could be multiple
instances of one type of realm or the other, but the number is
physically limited by the number of interfaces on the NAT device.
o Before the NAT can operate on a given interface, an "address map"
has to be configured on it. The address map in [RFC4008] is
equivalent to the pool tables in the present document. Since just
one "address map" is configured per interface, this is the
equivalent of a single address pool per interface.
o The address binding and port binding tables are roughly equivalent
to the address map and port map tables in the present document in
their content, but they can be either unidirectional or
bidirectional. The model in [RFC4008] shows the address binding
and port binding as alternative precursors to session
establishment, depending on whether the device does address
translation only or address and port translation. In contrast,
NATV2-MIB assumes a model where bidirectional port mappings are
based on bidirectional address mappings that have conceptually
been established beforehand.
o The equivalent to an [RFC4008] session in NATV2-MIB would be a
pair of port map entries. The added complexity in [RFC4008] is
due to the modeling of NAT service types as defined in [RFC3489]
(the symmetric NAT in particular) instead of the more granular set
of behaviors described in [RFC4787]. (Note: [RFC3489] has been
obsoleted by [RFC5389].)
With regard to that last point, the mapping between [RFC3489] service
types and [RFC4787] NAT behaviors is as follows:
o A full cone NAT exhibits endpoint-independent port mapping
behavior and endpoint-independent filtering behavior.
o A restricted cone NAT exhibits endpoint-independent port mapping
behavior, but address-dependent filtering behavior.
o A port restricted cone NAT exhibits endpoint-independent port
mapping behavior, but address-and-port-dependent filtering
o A symmetric NAT exhibits address-and-port-dependent port mapping
and filtering behaviors.
Note that these NAT types are a subset of the types that could be
configured according to the [RFC4787] behavioral classification used
in NATV2-MIB, but they include the two possibilities (full and
restricted cone NAT) that satisfy requirements REQ-1 and REQ-8 of
[RFC4787]. Note further that other behaviors defined in [RFC4787]
are not considered in [RFC4008].
Having established a context for discussion, we are now in a position
to compare the outputs provided to management from the [RFC4008] and
NATV2-MIB modules. This comparison relates to the ability to compare
results if testing with both MIBs implemented on the same device
during a transition period.
[RFC4008] provides three counters: incoming translations, outgoing
translations, and discarded packets, at the granularities of
interface, address map, and protocol, and incoming and outgoing
translations at the levels of individual address bind, address port
bind, and session entries. Implementation at the protocol and
address map levels is optional. NATV2-MIB provides a single total
(both directions) translations counter at the instance, protocol
within instance, and subscriber levels. Given the differences in
granularity, it appears that the only comparable measurement of
translations between the two MIB modules would be through aggregation
of the [RFC4008] interface counters to give a total number of
translations for the NAT instance.
NATV2-MIB has broken out the single discard counter into a number of
different counters reflecting the cause of the discard in more
detail, to help in troubleshooting. Again, with the differing levels
of granularity, the only comparable statistic would be through
aggregation to a single value of total discards per NAT instance.
Moving on to state variables, [RFC4008] offers counts of number of
"address map" (i.e., address pool) entries used (excluding static
entries) at the address map level and number of entries in the
address bind and address and port bind tables, respectively.
Finally, [RFC4008] provides a count of the number of sessions
currently using each entry in the address and port bind table. None
of these counts are directly comparable with the state values offered
by NATV2-MIB, because of the exclusion of static entries at the
address map level, and because of the differing models of the
translation tables between [RFC4008] and the NATV2-MIB.
6. Security Considerations
There are a number of management objects defined in this MIB module
with a MAX-ACCESS clause of read-write. Such objects may be
considered sensitive or vulnerable in some network environments. The
support for SET operations in a non-secure environment without proper
protection opens devices to attack. These are the tables and objects
and their sensitivity/vulnerability:
Limits: An attacker setting a very low or very high limit can easily
cause a denial-of-service situation.
Notification thresholds: An attacker setting an arbitrarily low
threshold can cause many useless notifications to be generated
(subject to the notification interval). Setting an arbitrarily
high threshold can effectively disable notifications, which could
be used to hide another attack.
Notification intervals: An attacker setting a low notification
interval in combination with a low threshold value can cause many
useless notifications to be generated.
Some of the readable objects in this MIB module (i.e., objects with a
MAX-ACCESS other than not-accessible) may be considered sensitive or
vulnerable in some network environments. It is thus important to
control even GET and/or NOTIFY access to these objects and possibly
to even encrypt the values of these objects when sending them over
the network via SNMP. These are the tables and objects and their
Objects that reveal host identities: Various objects can reveal the
identity of private hosts that are engaged in a session with
external end nodes. A curious outsider could monitor these to
assess the number of private hosts being supported by the NAT
device. Further, a disgruntled former employee of an enterprise
could use the information to break into specific private hosts by
intercepting the existing sessions or originating new sessions
into the host. If nothing else, unauthorized monitoring of these
objects will violate individual subscribers' privacy.
* entries in the natv2SubscriberTable;
* entries in the natv2AddressMapTable;
* entries in the natv2PortMapTable.
Other objects that reveal NAT state: Other managed objects in this
MIB may contain information that may be sensitive from a business
perspective, in that they may represent NAT capabilities, business
policies, and state information.
There are no objects that are sensitive in their own right, such as
passwords or monetary amounts.
SNMP versions prior to SNMPv3 did not include adequate security.
Even if the network itself is secure (for example by using IPsec),
there is no control as to who on the secure network is allowed to
access and GET/SET (read/change/create/delete) the objects in this
Implementations SHOULD provide the security features described by the
SNMPv3 framework (see [RFC3410]), and implementations claiming
compliance to the SNMPv3 standard MUST include full support for
authentication and privacy via the User-based Security Model (USM)
[RFC3414] with the AES cipher algorithm [RFC3826]. Implementations
MAY also provide support for the Transport Security Model (TSM)
[RFC5591] in combination with a secure transport such as SSH
[RFC5592] or TLS/DTLS [RFC6353].
Further, deployment of SNMP versions prior to SNMPv3 is NOT
RECOMMENDED. Instead, it is RECOMMENDED to deploy SNMPv3 and to
enable cryptographic security. It is then a customer/operator
responsibility to ensure that the SNMP entity giving access to an
instance of this MIB module is properly configured to give access to
Bantian, Longgang District
7100-8 Kit Creek Road
Research Triangle Park, North Carolina 27709
Phone: +1 919 392 5158
PT Taylor Consulting