Network Address Translation Overview

 

 

This chapter provides an overview of Network Address Translation (NAT) in-line service feature.

The following topics are covered in this chapter:

•	Supported Platforms and Products

•

Licenses

•	Supported Standards

•	NAT Feature Overview

•	How NAT Works

 

Supported Platforms and Products

NAT is an in-line service feature supported on the Cisco ASR 5000 chassis running 3GPP, 3GPP2, and LTE core network services (PDSN, HA, GGSN, and P-GW).

Important: For information on ASR 5000, please refer to the Product Overview Guide.

 

Licenses

NAT is a licensed in-line service feature requiring the following licenses:

•	Cisco PID [ ASR5K-00-CS01NAT ] NAT 1K Sessions with DPI, or Starent part number [ 600-00-7805 ] NAT/PAT With DPI.

•	Any other in-line service counting license (Enhanced Charging Service, Stateful Firewall, Content Filtering, etc.). For more information, please contact your local sales representative.

Important: For information on additional licenses required for enhanced or customer-specific features contact your local sales representative.

Important: For information on obtaining and installing licenses, refer to the Managing License Keys section of the Software Management Operations chapter in the System Administration Guide.

 

Supported Standards

The NAT feature supports the following RFCs:

•	RFC 1631: The IP Network Address Translator (NAT); May 1994

•	RFC 1918: Address Allocation for Private Internets; February 1996

•	RFC 2663: IP Network Address Translator (NAT) Terminology and Considerations; August 1999

•	RFC 2765: Stateless IP/ICMP Translation Algorithm (SIIT); February 2000

•	RFC 2766: Network Address Translation - Protocol Translation (NAT-PT); February 2000

•	RFC 3022: Traditional IP Network Address Translator (Traditional NAT); January 2001

•	RFC 3027: Protocol Complications with the IP Network Address Translator; January 2001

•	RFC 4787: Network Address Translation (NAT) Behavioral Requirements for Unicast UDP; January 2007

•	RFC 4966: Reasons to Move the Network Address Translator - Protocol Translator (NAT-PT) to Historic Status; July 2007

•	RFC draft-nishitani-cgn-00.txt: Carrier Grade Network Address Translator (NAT) Behavioral Requirements for Unicast UDP, TCP and ICMP; July 2, 2008

•	RFC draft-van-beijnum-behave-ftp64-06.txt: IPv6-to-IPv4 translation FTP considerations; May 19, 2009

•	RFC draft-ieftf-behave-dns64-11.txt: DNS64; February 15, 2010

•	RFC draft-ietf-behave-v6v4-xlate-stateful-12.txt: Stateful NAT64; July 10, 2010

•	RFC draft-ietf-behave-address-format-10.txt: IPv6 Addressing of IPv4/IPv6 Translators; August 16, 2010

•	RFC draft-ietf-behave-v6v4-framework-10.txt: Framework for IPv4/IPv6 Translation; August 17, 2010

•	RFC draft-ieft-behave-v6v4-xlate-23.txt: IP/ICMP Translation Algorithm; September 18, 2010

•	RFC 6052: IPv6 Addressing of IPv4/IPv6 Translators; October 2010

 

NAT Feature Overview

This section provides an overview of the NAT in-line service feature.

NAT translates non-routable private IP address(es) to routable public IP address(es) from a pool of public IP addresses that have been designated for NAT. This enables to conserve on the number of public IP addresses required to communicate with external networks, and ensures security as the IP address scheme for the internal network is masked from external hosts, and each outgoing and incoming packet goes through the translation process.

NAT works by inspecting both incoming and outgoing IP datagrams and, as needed, modifying the source IP address and port number in the IP header to reflect the configured NAT address mapping for outgoing datagrams. The reverse NAT translation is applied to incoming datagrams.

NAT can be used to perform address translation for simple IP and mobile IP. NAT can be selectively applied/denied to different flows (5-tuple connections) originating from subscribers based on the flows' L3/L4 characteristics—Source-IP, Source-Port, Destination-IP, Destination-Port, and Protocol.

Important: NAT works only on flows originating internally. Bi-directional NAT is not supported.

Important: NAT is supported only for TCP, UDP, and ICMP flows. For other flows NAT is bypassed. For GRE flows, NAT is supported only if the PPTP ALG is configured. For more information on ALGs, please refer to the NAT Application Level Gateway section.

Important: If a subscriber is assigned with a public IP address, NAT is not applied.

Important: To get NATed, the private IP addresses assigned to subscribers must be from the following ranges: Class A 10.0.0.0 – 10.255.255.255, Class B 172.16.0.0 – 172.31.255.255, and Class C 192.168.0.0 – 192.168.255.255

NAT supports the following mappings:

•	One-to-One: In one-to-one NAT each private IP address is mapped to a unique public NAT IP address. The private source ports do not change.

When a private IP address (IP1:port1) is mapped to a public IP address (IP2:port1), any packets from IP1:port1 will be sent as though via IP2:port1. The external host can only send packets to IP2:port1, which are translated to IP1:port1. The NAT port number will be the same as the source private port.

•

Many-to-One: In many-to-one NAT, multiple private IP addresses are mapped to a single public NAT IP address. In order to distinguish between different subscribers and different connections originating from same subscriber, internal private L4 source ports are translated to pre-assigned L4 NAT ports. Ports are allocated in chunks such that each private IP address is reserved a set of ports for future use. This is also known as Network Address Port Translation (NAPT).

Once a flow is marked to use a specific NAT IP address the same NAT IP address is used for all packets originating on that flow. The NAT IP address is released only when all flows and subscribers associated with it are released.

When all NAT IP addresses are in use, and a subscriber with a private IP address fails to get a NAT IP address for a specific flow, that specific flow will not be allowed and will fail.

All downlink—inbound from external networks—IP packets that do not match one of the existing NAT bindings are discarded by the system.

 

NAT Realms

A NAT realm is a pool of unique public IP addresses available for translation from private source IP addresses. IP addresses in a NAT IP pool are contiguous, and assignable as a subnet or a range that constitutes less than an entire subnet. IP addresses configured in NAT IP pools within a context must not overlap. At any time, within a context, a NAT IP address must be configured in any one NAT IP pool. IP addresses can be added to a NAT IP pool as a range of IP addresses.

Important: The minimum number of public IP addresses that must be allocated to each NAT IP pool must be greater than or equal to the number of Session Managers (SessMgrs) available on the system. On the ASR 5000, it is >= 84 public IP addresses. This can be met by a range of 84 host addresses from a single Class C. The remaining space from the Class C can be used for other allocations. Each address has available its port range ~64K ports.

Up to 2000 unique “IP pools + NAT IP pools” can be configured per context. A maximum of three NAT IP pools/NAT IP pool groups can be configured in a Firewall-and-NAT policy. At any time a subscriber can be associated with a maximum of three different NAT IP pools/NAT IP pool groups and can have NATed flows on three different NAT IP addresses at the same time.

Allocation of NAT IP addresses in NAT IP pools to subscriber traffic is based on the L3/L4 characteristics—IP addresses, ports, and protocol—of the subscriber flows. It is possible to configure the system to perform or not perform NAT based on one or more L3/L4 parameters. This feature is also known as Target-based NAT. For more information, see the Target-based NAT Configuration section.

NAT IP pools have the following configurable parameters. These parameters are applicable to all IP addresses in a NAT IP pool.

•	NAT IP Address Allocation Mode: Specifies when to allocate a NAT IP address to a subscriber; either at call setup or during data flow based on the allocation mode.

•

Not-on-demand Allocation Mode: This is the default mode. In this mode, the NAT IP address is allocated to the subscriber at call setup. If there are three NAT IP pools/NAT IP pool groups (maximum possible) configured in the subscriber’s Firewall-and-NAT policy, the subscriber is allocated three NAT IP addresses, one from each NAT IP pool/NAT IP pool group based on rule matching.

•	On-demand Allocation Mode: In this mode NAT resources are assigned and allocated dynamically based on subscriber flows. The NAT IP address is allocated to the subscriber when the data traffic flows in and not at call setup.

In case of on-demand pools, since the NAT IP address is not allocated to the subscriber at call setup, the subscriber may not have a NAT IP address allocated when the first packet is received. Until the successful allocation of a NAT IP address, based on the configuration, the packets can either be buffered or dropped. Once a free NAT IP address is available, it is allocated to the subscriber to be used for flows matching the pool.

•	NAT Binding Timer: Specifies the timeout period, in seconds, to deallocate NAT resources that were allocated to subscriber flows. When a subscriber flow stops the timer starts counting down, and on expiry the NAT resources are deallocated to be made available for other subscriber flows.

•	In one-to-one allocation, for a given NAT IP address, the NAT Binding Timer starts counting down when there are no active flows using that NAT IP address. When the NAT Binding Timer expires, the NAT IP address gets deallocated.

•

In many-to-one allocation, wherein subscribers are allocated port-chunks rather than individual ports, as long as a port-chunk is allocated to a subscriber, all ports from that port-chunk are reserved for that subscriber. When all flows using ports from that port-chunk get timed out/cleared, the NAT Binding Timer starts counting down. If any new flows come up before the NAT Binding Timer expires, ports are once again allocated from that port-chunk, and the NAT Binding Timer gets cancelled. As long as there are active flows using the port-chunk it cannot be deallocated. But, if no new flows come and the NAT Binding Timer expires, the port-chunk gets deallocated. In the case of on-demand NAT, if it is the last port-chunk for the NAT IP address, on NAT Binding Timer expiry, the NAT IP address gets deallocated along with the last port-chunk.

•	Maximum Users per NAT IP Address: Applicable only to many-to-one NAT IP pools. Specifies the maximum number of subscribers sharing one NAT IP address. A maximum of 2016 subscribers can be configured per NAT IP address.

•	Port Chunk Size: Applicable only to many-to-one NAT IP pools. Specifies the block size of contiguous ports to be assigned to a many-to-one NAT subscriber. This number has to be divisible by 32 up to a maximum of 32,256.

•

Maximum Port-chunks per User: Applicable only to many-to-one NAT IP pools. Specifies the maximum number of port-chunks allowed for an individual subscriber from the same NAT IP address. This will limit subscribers from dominating all the available ports in a many-to-one NAT IP. A maximum of 2016 port-chunks can be configured per subscriber.

Consider a case where a single TCP flow is active in a port-chunk. When this connection gets cleared, the TCP NAT port goes to Time Wait state. Since it is the last flow of the port-chunk, the NAT Binding Timer also gets started. Assume NAT Binding Timer >= TCP 2MSL Timer. Once the 2MSL Timer expires, the TCP port would go to Free state. However, the NAT Binding Timer keeps running. On NAT Binding Timer expiry, the port-chunk is deallocated. If this was the last port-chunk for that subscriber, the NAT IP address is also deallocated along with this port-chunk.

In case NAT Binding Timer < TCP 2MSL Timer, at NAT Binding Timer expiry, the TCP port is forcefully moved to Free state from Time Wait state and the port-chunk deallocated.

•	Port Chunk Thresholds: Applicable only to many-to-one NAT IP pools. Specifies threshold in terms of percentage of allocated port-chunks against total port-chunks available. Once the threshold is reached, new subscribers will not be allocated the same NAT IP address.

•	AAA Binding Update Message Required: Applicable only to one-to-one NAT IP pools. Enables AAA binding messages for one-to-one NAT IP pools. This is not supported for many-to-one NAT IP pools.

•	Alert Thresholds: Threshold limits can be specified to trigger alarms for NAT IP pools for pool-used, pool-free, pool-hold, and pool-release cases.

•	SRP-Activate: Applicable to both one-to-one and many-to-one NAT IP pools. When configured, the NAT IP pool will become usable only when the SRP state is active.

 

NAT IP Pool Groups

Similar NAT IP pools can be grouped into NAT IP pool groups. This enables to bind discontiguous IP address blocks in individual NAT IP pools to a single NAT IP pool group.

When configuring a NAT IP pool group, note that only those NAT IP pools that have similar characteristics can be grouped together. The similarity is determined by the NAT IP pool Type (One-to-One / Many-to-One), users configured per NAT IP address (applicable only to many-to-one NAT IP pools), NAT IP Address Allocation Mode (On-demand/Not-on-demand), and Port Chunk Size (applicable only to many-to-one NAT IP pools) parameters. Dissimilar NAT IP pools cannot be grouped together.

It is recommended that all the NAT IP pools in a NAT IP pool group be configured with the same values for the other parameters, so that the NAT behavior is predictable across all NAT IP pools in that NAT IP pool group.

The NAT IP pool from which a NAT IP address is assigned will determine the actual values to use for all parameters.

It is recommended that in a Firewall-and-NAT policy all the realms configured either be NAT IP pools or NAT IP pool groups. If both NAT IP pool(s) and NAT IP pool group(s) are configured, ensure that none of the NAT IP pool(s) are also included in the NAT IP pool group.

 

NAT IP Address Allocation and Deallocation

Cisco System’s implementation of NAPT is Endpoint-independent Mapping, wherein NAT reuses the same NAT source port mapping for subsequent packets sent from the same private IP address and port, and with the same protocol to any public destination host IP address and port.

That is, all flows coming from the subscriber for the current session with the same protocol and same source IP address and source port (X:x) would get the same NAT IP address and NAT port (X:x) irrespective of the destination IP address and port. NAT will not allow any inbound packets to the NAT IP address and NAT port (X:x) from an external host IP address and host port (Y:y), unless the internal host (MS) had previously sent a packet of the same protocol type to that external IP address and Port (Y:y). However, this behavior changes if NAT ALG is enabled. The ALG creates pin holes / dynamic routes in the NAT and allows downlink packets that match the pin holes / dynamic routes towards the internal host (MS) given that there was already a parent connection from MS towards the external host.

The advantage of endpoint-independent mapping is that applications are unaffected by NAT translations.

Inbound connection to the NAT IP address can be allowed in one-to-one pools based on configuration.

 

NAT IP Address Allocation

The NAT IP address is allocated based on the following parameters:

•	Maximum Users per NAT IP Address: The maximum number of subscribers sharing a NAT IP address. Once the number of active subscribers using a NAT IP address reaches this limit, that NAT IP address will not be allocated to new subscribers.

•

Port-chunk Thresholds: The threshold is configured in percentage of total number of port-chunks. If the number of port-chunks already allocated from a given NAT IP address is less than the configured threshold limit of port-chunks, then the NAT IP address can be chosen for a new subscriber provided the “Maximum Users per NAT IP Address” is not reached. But if the number of chunks allocated is greater than or equal to the threshold limit of port-chunks, then the NAT IP address will not be chosen for a new subscriber. The remaining free port-chunks will be used for existing subscribers using the NAT IP address.

 

NAT IP Address Deallocation

Whenever a NAT IP address is deallocated, all the port-chunks associated with the subscriber are released back to the pool.

In case there is only one port-chunk associated with the subscriber:

•	In case of many-to-one not-on-demand NAT IP pools, the last port-chunk is not released back to the pool even after NAT Binding Timer expires. Only when the call gets disconnected, the port-chunk is released along with the NAT IP address.

•	In case of many-to-one on-demand NAT IP pools, when the last flow using the port-chunk gets cleared, the NAT Binding Timer is started. When the NAT Binding Timer expires, the port-chunk along with the NAT IP address is released back to the pool.

•	In case of one-to-one on-demand NAT IP pools, when there are no active flows using a NAT IP address, the NAT Binding Timer is started. When the NAT Binding Timer expires, the NAT IP address gets deallocated.

 

NAT Port-chunk Allocation and Deallocation

This section describes the Port-chunk Allocation and Deallocation feature for many-to-one NAT.

 

NAT Port-chunk Allocation

Subscribers sharing a NAT IP address are allocated NAT ports in chunks. The ports in a port-chunk are always used for the subscriber to whom that port-chunk is allocated irrespective of the protocol.

Whenever a NAT IP address gets allocated to a subscriber, the first port-chunk gets allocated along with the NAT IP address. Thus, for not-on-demand pools, the first port-chunk gets allocated during call setup, and for on-demand pools during data flow.

A subscriber’s TCP and UDP data traffic is NATed with ports chosen in a random fashion from the port-chunk allocated to that subscriber. For other protocol traffic, the first available port is allocated. When all the ports in a port-chunk are in use, a free port-chunk is requested for. A new port-chunk is only allocated if the “Maximum Port-chunks Per User” limit is not reached.

 

NAT Port-chunk Deallocation

A port-chunk gets deallocated in the following cases:

•	“NAT Binding Timer” expiry

•	Subscriber session disconnect

 

NAT Binding Timer

When all flows using ports from a particular port-chunk get timed out/cleared, the port-chunk gets freed. When the last port of that port-chunk gets freed, the NAT Binding Timer starts counting. Before the NAT Binding Timer expires, if any new flows come up, ports are reallocated from the port-chunk, and the timer gets cancelled. The port-chunk cannot be deallocated as long as there are active flows using that port-chunk. But, if no new flows come and the NAT Binding Timer expires, the port-chunk gets deallocated.

In case of not-on-demand pools, the additional port-chunks that were allocated on demand will be deallocated based on the NAT binding timeout. However, the last port-chunk will not be deallocated even after the Binding Timer expires. This last port-chunk will only be deallocated when the NAT IP address is deallocated from the subscriber.

In case of on-demand pools, the port-chunks are deallocated based on the NAT binding timeout. When the last port-chunk gets freed, the NAT IP address also gets deallocated from the subscriber.

It is ensured that a port-chunk is associated with the subscriber as long as a valid NAT IP address is allocated to the subscriber.

 

Subscriber Session Disconnect

When a subscriber disconnects, all port-chunks associated with that subscriber are freed.

If the NAT Binding Timer has not expired, the port-chunks will not be usable immediately, only on NAT Binding Timer expiry will the port-chunks become available for new subscribers.

 

NAT IP Address/Port Allocation Failure

When a packet cannot be translated, the application can be notified by way of ICMP error messages, if configured. Translation failures may be due to no NAT IP address or port being available for translation.

Important: In the case of P-GW, NAT IP Address/Port Allocation Failure notification is not applicable.

 

TCP 2MSL Timer

NAT does port management only for many-to-one pools. Hence, The TCP 2MSL timer is only available for many-to-one NAT. It is necessary to ensure that a TCP NAT port in Time Wait state is not reused if there are other free ports available for the subscriber. If such a reuse happens, then there is a possibility that connections might get terminated by the server. To avoid such issues, whenever a many-to-one NAT TCP flow gets cleared, the NAT port goes to Time Wait state (2MSL started for that port). Once 2MSL timer expires, the NAT port becomes usable. The 2MSL timer is started for every TCP NAT port as soon as the TCP connection gets cleared. This ensures that a NAT TCP port gets reused only after expiry of the configured TCP 2MSL timer.

Consider a case where a single TCP flow is active in a port-chunk. When this connection gets cleared, the TCP NAT port goes to Time Wait state. Since this is the last flow of the port-chunk, the NAT Binding Timer also gets started. Assume NAT Binding timer >= TCP 2MSL timer. Once the 2MSL timer expires, the TCP port becomes usable. However, the NAT Binding Timer keeps counting, and on expiry, the port-chunk is released.

In case the NAT Binding Timer < TCP 2MSL Timer, on NAT Binding Timer expiry, the TCP port is forcefully moved to Free State (made usable) from Time Wait state and the port-chunk released.

 

Flow Mapping Timer

The Flow Mapping timer is a new timer implemented as an extension to the existing idle-timeout in ECS, and is supported only for TCP and UDP flows. This flow mapping applies only for NAT enabled calls.

The purpose of this timer is to hold the resources such as NAT IP, NAT port, and Private IP NPU flow associated with a 5-tuple ECS flow until Mapping timeout expiry. If the feature is disabled, the Flow mapping timeout will not get triggered for TCP/UDP idle timed out flows. The resources such as NAT mapping will be released with the 5-tuple flow itself.

 

NAT Binding Records

Whenever a NAT IP address or NAT port-chunk is allocated/deallocated to/from a subscriber, NAT Binding Records (NBR) can be generated. Generation of NBRs is configurable in the Firewall-and-NAT policy configuration.

Important: NAT Binding Records are not supported for NAT64 in this release.

NBRs are supported for both on-demand and not-on-demand NAT IP pools. For a one-to-one NAT IP pool, an NBR is generated whenever a NAT IP address is allocated/deallocated to/from a subscriber. For a many-to-one NAT IP pool, an NBR is generated when a port-chunk is allocated/deallocated to/from a subscriber for a NAT IP address. It is also possible to configure generation of NBRs only when a port-chunk is allocated, or deallocated, or in both cases.

The following is the list of attributes that can be present in NBRs. You can configure a subset of these attributes or all of them to be logged in NBRs. If an attribute is not available, while logging records that field is populated with NULL.

•	ip subscriber-ip-address: The private IP address

•	radius-calling-station-id: The IMSI of the mobile node.

•	radius-fa-nas-identifier: A string that identifies PDSN. This field is optional if PDSN-NAS-IP address field is present.

•	radius-fa-nas-ip-address:

•	radius-user-name: NAI of the mobile node.

•	sn-correlation-id: If available. The HA-Correlation-ID identifying the entire MIP session.

•	sn-fa-correlation-id: If available. The PDSN-Correlation-ID as sent by the PDSN using the same format and length.

•	sn-nat-binding-timer: Optional. The NAT Binding Timer assigned to the Realm.

•	sn-nat-gmt-offset: Optional. The offset from GMT to correlate timestamps of records; GMT offset of the node generating this record. For example: -5.00, +5.30

•	sn-nat-ip: The NAT IP address of mobile node.

•	sn-nat-last-activity-time-gmt: The time the last flow in a specific NAT set of flows was seen in GMT time.

•	sn-nat-port-block-end: The NAT Port Block End of the mobile node.

•	sn-nat-port-block-start: The NAT Port Block Start of the mobile node.

•	sn-nat-port-chunk-alloc-dealloc-flag: 1: allocate; 0: deallocate

•	sn-nat-port-chunk-alloc-time-gmt: The NAT Port Chunk Allocation Timestamp (Sample time format: 03/11/2009 10:38:35)

•	sn-nat-port-chunk-dealloc-time-gmt: The NAT Port Chunk Deallocation Timestamp (Sample time format: 03/11/2009 10:38:35)

•	sn-nat-realm-name: Optional. The name of the locally configured NAT Realm.

•	sn-nat-subscribers-per-ip-address: Optional. NAT Multiplier assigned to the Realm.

•	bearer 3gpp charging-id: The charging ID for the PDN Session.

•	bearer 3gpp sgsn-address: The SGW/SGSN address.

•	bearer ggsn-address: The PGW/GGSN address.

•	bearer 3gpp imsi: The IMSI value of the subscriber.

Important: The NBR attributes: sn-correlation-id, sn-fa-correlation-id, radius-fa-nas-ip-address, radius-fa-nas-identifier are not applicable for PGW and GGSN.

 

NAT Binding Updates

Whenever a NAT IP address or NAT port-chunk is allocated/deallocated to/from a subscriber, to update NAT binding information for that subscriber in the AAA, a NAT Binding Update (NBU) can be sent to the AAA server.

Important: NAT Binding Updates are not supported for NAT64 in this release.

Important: In this release, P-GW and GGSN do not support the NBU feature.

Since port-chunk allocation/deallocation happens on a per-call basis, this ensures that AAA messaging is reduced to a great extent. NBUs are sent to the AAA server in accounting-interim messages. To send or not to send NBUs to the AAA server is configurable in the NAT IP pool configuration.

NBUs are supported for both one-to-one and many-to-one NAT IP pools.

An NBU contains the following attributes:

•	Alloc-Flag

•	Binding-Timer

•	Correlation-Id

•	Loading-Factor

•	NAT-IP-Address

•	NAT-Port-Block-End: In the case of one-to-one NAT, the value is 65535

•	NAT-Port-Block-Start: In the case of one-to-one NAT, the value is 1

 

CoA NAT Query

If the NAT binding information is not available at the AAA, the AAA server can query the chassis for the information. This query uses the Change of Authorization (CoA) format, wherein the AAA sends a one-to-one NAT IP address as a query, and in the CoA query response the NBU is obtained if available at the time of query.

Important: CoA NAT Query is not supported for NAT64 in this release.

Important: In this release, CoA query for NAT binding information is only supported for one-to-one NAT.

The CoA query request must contain the following attributes:

•	Event-Timestamp

•	NAS-IP-Address

•	SN1-NAT-IP-Address

Important: For SN1-NAT-IP-Address, this release supports VSA-Type values 0 and 1.

For a successful query, the CoA ACK response contains the following attributes:

•	Acct-Session-Id

•	Correlation-Id

•	Framed-IP-Address

•	NAT-IP-Address

•	NAT-Port-Block-End

•	NAT-Port-Block-Start

•

User-Name

Important: For information on the AVPs/VSAs, please refer to the AAA and GTPP Interface Administration and Reference.

 

Firewall-and-NAT Policy

Firewall-and-NAT policies are configured in the CLI Firewall-and-NAT Policy Configuration Mode. Each policy contains a set of access ruledefs with priorities and actions, and the NAT configurations. On a system, multiple such policies can be configured, however at any point of time only one policy is associated to a subscriber.

Important: In release 8.x, NAT for CDMA and early UMTS releases used rulebase-based configurations, whereas in later UMTS releases NAT used policy-based configurations. In 9.0 and later releases, NAT for UMTS and CDMA releases both use policy-based configurations. For more information, please contact your local service representative.

Important: In a Firewall-and-NAT policy, a maximum of three NAT IP pools/NAT IP pool groups can be configured. A subscriber can be allocated only one NAT IP address per NAT IP pool/NAT IP pool group, hence at anytime, there can only be a maximum of three NAT IP addresses allocated to a subscriber.

New NAT IP pools/NAT IP pool groups cannot be added to a policy if the maximum allowed is already configured in it. However, a pool/pool group can be removed and then a new one added. When a pool/pool group is removed and a new one added, the pool/pool group that was removed will stay associated with the subscriber as long as the subscriber has active flows using that pool/pool group. If the subscriber is already associated with three NAT IP pools (maximum allowed), any new flows from that subscriber for the newly added pool will be dropped. A deleted pool is disassociated from the subscriber on termination of all flows from that subscriber using that pool. The new pool/pool group is associated with the subscriber only when the subscriber sends a packet to the newly added pool.

In the Firewall-and-NAT policy configuration, the NAT44/NAT64 policy must be enabled. Once NAT is enabled for a subscriber, the NAT IP address to be used is chosen from the NAT IP pools/NAT IP pool groups specified in matching access rules configured in the Firewall-and-NAT policy.

The Firewall-and-NAT policy used for a subscriber can be changed either from the command line interface, or through dynamic update of policy name in Diameter and RADIUS messages. In both the cases, NAT status on the active call remains unchanged.

The Firewall-and-NAT policy to be used for a subscriber can be configured in:

•	ECS Rulebase: The default Firewall-and-NAT policy configured in the ECS rulebase has the least priority. If there is no policy configured in the APN/subscriber template, and/or no policy to use is received from the AAA/OCS, only then the default policy configured in the ECS rulebase is used.

•

APN/Subscriber Template: The Firewall-and-NAT policy configured in the APN/subscriber template overrides the default policy configured in the ECS rulebase. To use the default policy configured in the ECS rulebase, in the APN/subscriber configuration, the command to use the default rulebase policy must be configured.

•	AAA/OCS: The Firewall-and-NAT policy to be used can come from the AAA server or the OCS. If the policy comes from the AAA/OCS, it will override the policy configured in the APN/subscriber template and/or the ECS rulebase.

Important: The Firewall-and-NAT policy received from the AAA and OCS have the same priority. Whichever comes latest, either from AAA/OCS, is applied.

The Firewall-and-NAT policy to use can also be received from RADIUS during authentication.

 

Disabling NAT Policy

Important: By default, NAT processing for subscribers is disabled.

NAT processing for subscribers is disabled in the following cases:

•	If the AAA/OCS sends the SN-Firewall-Policy AVP with the string “disable”, the locally configured Firewall-and-NAT policy does not get applied.

•	If the SN-Firewall-Policy AVP is received with the string “NULL”, the existing Firewall-and-NAT policy will continue.

•	If the SN-Firewall-Policy AVP is received with a name that is not configured locally, the subscriber session is terminated.

 

Updating Firewall-and-NAT Policy in Mid-session

The Firewall-and-NAT policy can be updated mid-session provided the policy was enabled during call setup.

Important: When the firewall AVP contains “disable” during mid-session firewall policy change, there will be no action taken as the Firewall-and-NAT policy cannot be disabled dynamically. The policy currently applied will continue.

Important: For all NAT-enabled subscribers, when the Firewall-and-NAT policy is deleted, the call is dropped.

In a Firewall-and-NAT policy, you can change the NAT enabled/disabled status at any time. However, the updated NAT status will only be applied to new calls, active calls using that Firewall-and-NAT policy will remain unaffected.

 

Target-based NAT Configuration

A NAT IP pool can be selected based on the L3/L4 characteristics of a subscriber’s flows. NAT can be configured such that all subscriber traffic coming towards specific public IP address(es) always selects a specific NAT IP pool based on the L3/L4 traffic characteristics.

Important: A subscriber can be allocated only one NAT IP address per NAT IP pool/NAT IP pool group from a maximum of three NAT IP pools/NAT IP pool groups. Hence, at anytime, there can only be a maximum of three NAT IP addresses allocated to a subscriber.

This association is done with the help of access ruledefs configured in the Firewall-and-NAT policy. The NAT IP pool/NAT IP address to be used for a subscriber flow is decided during rule match. When packets match an access ruledef, NAT is applied using the NAT IP address allocated to the subscriber from the NAT IP pool/NAT IP pool group configured in that access ruledef.

If no NAT IP pool/NAT IP pool group name is configured in the access ruledef matching the packet, and if there is a NAT IP pool/NAT IP pool group configured for “no ruledef matches”, a NAT IP address from the NAT IP pool/NAT IP pool group configured for “no ruledef matches” is allocated to the flow.

If no NAT IP pool/NAT IP pool group is configured for “no ruledef matches” and if there is a default NAT IP pool/NAT IP pool group configured in the rulebase, a NAT IP address from this default NAT IP pool/NAT IP pool group is allocated to the flow.

If a NAT IP pool/NAT IP pool group is not configured in any of the above cases, no NAT will be performed for the flow. Or, if bypass NAT is configured in a matched access rule or for “no ruledef matches” then NAT will not be applied even if the default NAT IP pool/NAT IP pool group is configured. The order of priority is:

1.	Bypass NAT

2.	NAT IP pool/NAT IP pool group in ruledef

3.	NAT IP pool/NAT IP pool group for “no-ruledef-matches”

4.	Default NAT IP pool/NAT IP pool group

When a new NAT IP pool/NAT IP pool group is added to a Firewall-and-NAT policy, it is associated with the active subscriber (call) only if that call is associated with less than three (maximum limit) NAT IP pools/NAT IP pool groups. If the subscriber is already associated with three NAT IP pools/NAT IP pool groups, any new flows referring to the newly added NAT IP pool/NAT IP pool group will get dropped. The newly added NAT IP pool/NAT IP pool group is associated to a call only when one of the previously associated NAT IP pools/NAT IP pool groups is freed from the call.

 

NAT Application Level Gateway

Some network applications exchange IP/port information of the host endpoints as part of the packet payload. This information is used to create new flows, by server or client.

As part of NAT ALGs, the IP/port information is extracted from the payload, and the flows are allowed dynamically (through pinholes). IP and port translations are done accordingly. However, the sender application may not be aware of these translations since these are transparent, so they insert the private IP or port in the payload as usual.

For example, FTP NAT ALG interprets “PORT” and “PASV reply” messages, and NAT translates the same in the payload so that FTP happens transparently through NAT. This payload-level translation is handled by the NAT ALG module.

The NAT module will have multiple NAT ALGs for each individual application or protocol.

 

Supported NAT ALGs

This release supports NAT ALGs only for the following protocols:

•

H323

•	File Transfer Protocol (FTP)

•	Point-to-Point Tunneling Protocol (PPTP): If PPTP ALG is enabled, NAT is supported for GRE flows that are generated by PPTP.

•	Real Time Streaming Protocol (RTSP)

•	Session Initiation Protocol (SIP)

•	Trivial File Transfer Protocol (TFTP)

For NAT ALG processing, in the rulebase, routing rules must be configured to route packets to the corresponding analyzers.

This release now supports session recovery for SIP ALG. Only one contact pinhole, and only one connected call and its associated media pinholes will be recovered for a subscriber. Any subscriptions, ongoing transactions, or unconnected calls will not be recovered. SIP ALG recovery data will be check-pointed using the variable length micro checkpointing mechanism.

 

H323 ALG Support

This release provides support for H323 ALG that is designed to traverse NAT by inspecting and altering information contained in existing H323 messages as they pass through the NAT. It can alter address and port information in registration, call signaling and automatically open pinholes in the NAT to allow media flow.

H323 ALG performs the following functions:

•	Communicates with the core for binding management

•	Uses H323 stack for parsing and encoding the H323 messages

•	Communicates with NAT for signaling messages

•	Performs protocol specific processing if required

The following supplementary services are currently supported in H323 ALG:

•	Call Transfer: The Call Transfer supplementary service enables the served user (User A) to transform an existing call with a User B (primary call) into a new call between current User B and a new User C (transferred-to user) selected by served user A.

•	Call Hold: The Call Hold supplementary service allows the served user, which may be the originally calling or the called user, to interrupt communications on an existing call and then subsequently, if desired, re-establish (i.e. retrieve) communications with the held user.

•

Call Diversion: Call Diversion supplementary service permits a served user to have incoming calls addressed to the served user's number redirected to another number; on busy service, it enables a served user to have calls redirected to another endpoint; on No Answer, it enables a served user to have calls addressed to the served endpoint's number and redirected to another endpoint if the connection is not established within a defined period of time.

•	Call Waiting: The Call Waiting supplementary service permits a busy user to be informed of an incoming call while being engaged with one or more other calls.

•	Call Offering: The Call Offering supplementary service on request from the calling user, enables a call to be offered to a busy user and to wait for that called user to accept the call, after the necessary resources have become available.

 

NAT Aware H323 Clients

An application layer gateway, at the Firewall/NAT, examines all the H323 packets and modifies the packet such that all the private addresses are replaced by public addresses. It also opens all the pinholes required for successful call establishment. A NAT aware endpoint establishes end-to-end media session through FW/NAT without the need of ALG. Any TCP connection or UDP packet sent from the internal network through the firewall opens a pinhole dynamically in the firewall. This pinhole allows incoming messages to be sent from the destination of the TCP connection or the UDP packet. The pinhole stays open as long as the network sends information through the pinhole to the same destination.

If an end point supports NAT traversal, H323 ALG disables itself so that end point directly opens required pinhole and establishes media path between them. The ALG will not manage any pinhole for media traversal across Firewall/NAT for NAT aware clients. By default, the ALG will bypass all the clients that support H460.18/19 and H460.23/24.

 

EDRs and UDRs

This section describes the NAT-specific attributes supported in EDRs and UDRs.

 

EDRs

The following NAT-specific attributes are supported in regular EDRs:

•	sn-nat-subscribers-per-ip-address: Subscriber(s) per NAT IP address

•	sn-subscriber-nat-flow-ip: NAT IP address of NAT-enabled subscribers

•	sn-subscriber-nat-flow-port: NAT port number of NAT-enabled subscribers

 

UDRs

The following NAT-specific attribute is supported in regular UDRs:

sn-subscriber-nat-flow-ip: NAT IP addresses that are being used by NAT-enabled subscribers. The NAT IP addresses assigned from each of the associated pool for the call are logged. A space is used as a separator between individual IP addresses.

 

Bulk Statistics

The NAT realms are configured in a context and statistics are stored per context per realm. These statistic variables, both cumulative and snapshot, are available in the nat-realm schema.

Bulkstats are only generated for the first 100 NAT IP pools from an alphabetical list of all NAT IP pools, which is based on the context name and pool name. Therefore, to generate bulkstats for a specific NAT IP pool it must be named such that it gets selected in the first 100 bulkstats.

The following are cumulative statistics that can be part of NAT bulkstats:

•	vpnname: Context name

•	realmname: Realm name

•	nat-bind-updates: Total interim AAA NBU sent.

•	nat-rlm-bytes-tx: Total number of NAT44 and NAT64 bytes transferred by realm (uplink + downlink).

•	nat-rlm-bytes-nat44-tx: Total number of NAT44 bytes transferred by realm.

•	nat-rlm-bytes-nat64-tx: Total number of NAT64 bytes transferred by realm.

•	nat-rlm-flows: Total number of NAT44 and NAT64 flows used by the realm.

•	nat-rlm-nat44-flows: Total number of NAT44 flows processed by realm.

•	nat-rlm-nat64-flows: Total number of NAT64 flows processed by realm.

•	nat-rlm-ip-denied: Total number of NAT44 and NAT64 flows denied NAT IP address.

•	nat-rlm-ip-denied-nat44: Total number of NAT44 flows denied IP.

•	nat-rlm-ip-denied-nat64: Total number of NAT64 flows denied IP.

•	nat-rlm-port-denied: Total number of NAT44 and NAT64 flows denied ports.

•	nat-rlm-port-denied-nat44: Total number of NAT44 flows denied ports.

•	nat-rlm-port-denied-nat64: Total number of NAT64 flows denied ports.

•	nat-rlm-max-port-chunk-subs: Total number of subscribers who used maximum number of port chunks.

•	nat-rlm-max-port-chunk-used: Maximum port chunks used.

The following are snapshot statistics that can be part of NAT bulkstats:

•	vpnname: Context name

•	realmname: Realm name

•	nat-rlm-ttl-ips: Total number of NAT public IP addresses, per context per NAT realm. Is a static value.

•	nat-rlm-ips-in-use: Total number of NAT IP addresses currently in use, per context per NAT realm.

•	nat-rlm-current-users: Total number of subscribers currently using the NAT realm.

•	nat-rlm-ttl-port-chunks: Total number port-chunks, per context per NAT realm. Is a static value.

•	nat-rlm-chunks-in-use: Total number of port-chunks currently in use, per context per NAT realm.

•	nat-rlm-max-cur-port-chunk-subs: Current number of subscribers using maximum number of port chunks.

•	nat-rlm-max-cur-port-chunk-used: Maximum port chunks used by active subscribers.

•	nat-rlm-port-chunk-size: Size of the port chunk in the NAT realm.

•	nat-rlm-port-chunk-average-usage-tcp: Average TCP port usage in the allocated TCP ports, i.e. out of allocated TCP ports how many got used. Not percentage value.

•	nat-rlm-port-chunk-average-usage-udp: Average UDP port usage in the allocated UDP ports, i.e. out of allocated UDP ports how many got used. Not percentage value.

•	nat-rlm-port-chunk-average-usage-others: Average other (ICMP or GRE) port usage in the allocated other ports, i.e. out of allocated ‘other’ ports how many got used. Not percentage value.

 

Alarms

Alert threshold values can be specified to generate alarms for NAT IP pools. To specify realm-specific threshold limits (pool-used, pool-free, pool-release, and pool-hold) “alert-threshold” NAT IP pool parameter can be used, or it can also be specified across context. These thresholds can be specified to any number of NAT IP pools.

In case of many-to-one NAT, it is possible to specify port-chunks usage threshold per NAT IP pool. This threshold value is applicable to all many-to-one NAT IP pools across the system. However, note that alarms are only generated for the first 100 many-to-one NAT IP pools from an alphabetical list of all NAT IP pools.

 

Session Recovery and ICSR

In session recovery, as part of the Private IP assigned to the subscriber:

•	The public IP address used for the subscriber is recovered. The NAT IP address being used by the subscriber can be on-demand or not-on-demand. In case of many-to-one NAT, the port-chunks associated with the NAT IP address for the subscriber needs to check-pointed as well.

•	In case Bypass NAT feature is used, then the private IP flow needs to be recovered.

To be recovered the NAT IP addresses need to be checkpointed. The checkpointing can be:

•	Full Checkpoint

•	Micro Checkpoint

To recover the bypass NAT flow, the bypass flow needs to be checkpointed. The checkpointing of Bypass NAT flow can be:

•	Full Checkpoint

•	Micro Checkpoint

In case of not-on-demand, the NAT IP address being used by the subscriber is known after call setup. This gets checkpointed as part of the normal full checkpoint. In case of on-demand NAT, the NAT IP address being used by the subscriber is known only in the data-path. This will be checkpointed as part of micro checkpoint.

In case of many-to-one NAT, the port-chunks being used will always be checkpointed as part of micro checkpoint.

In case of bypass NAT flow, in most cases the flow gets checkpointed as part of micro checkpoint.

Any information that is checkpointed as part of full checkpoint is always recovered. Data checkpointed through micro checkpoint cannot be guaranteed to be recovered. The timing of switchover plays a role for recovery of data done through micro checkpoint. If failover happens after micro checkpoint is completed, then the micro checkpointed data will get recovered. If failover happens during micro checkpoint, then the data recovered will be the one obtained from full checkpoint.

Once NAT IP/and Port-Chunks/Bypass NAT flow are recovered, the following holds good:

•	One-to-one NAT: Since NAT IP address being used for one-to-one NAT is recovered, on-going flows will be recovered as part of Firewall Flow Recovery algorithm as one-to-one NAT does not change the port.

•	Many-to-one NAT: On-going flows will not be recovered as the port numbers being used for flows across chassis peers/SessMgr peers are not preserved.

It is now possible to enable/disable the checkpointing of NATed flows and control the type of flows to be checkpointed based on criteria. Check pointing is done only for TCP and UDP flows.

Many-to-one NAT flow recovery is supported for ICSR in this release.

•	Bypass NAT Flow: On-going flows will be recovered as part of Firewall Flow Recovery algorithm.

All of the above items is applicable for ICSR as well. In this release, SIP ALG now supports ICSR and is applicable only to UDP flows.

Category	Event		Impacted	Details
One-to-One NAT	Session		No	Session recovered.
	New Traffic		No	NAT will be applied.
	Ongoing Traffic		Yes	Cannot differentiate between ongoing traffic and unsolicited traffic. A rule-match is done and if allowed, NAT will be applied accordingly on the packet.
	Unsolicited Traffic (downlink packets)		Yes	Cannot differentiate between ongoing traffic and unsolicited traffic. Translation will be done and packet action taken based on the rule-match.
Many-to-One NAT	Session		No	Session recovered.
	New Traffic		No	NAT will be applied.
	Ongoing Traffic	TCP	Yes	Packet will be dropped.
		UDP	Yes and No	If it is downlink packet, it will be dropped. If it is uplink packet, NAT will be applied with a new port.
		ICMP	Yes and No	If it is downlink packet, it will be dropped. If it is uplink packet, NAT will be applied with a new port.
	Unsolicited Traffic (downlink packets)		No	Packet will be dropped.
Bypass NAT	Session		No	Session recovered.
	New Traffic		No	Traffic will be NAT bypassed.
	Ongoing Traffic		No	Traffic will be NAT bypassed.
	Unsolicited Traffic (downlink packets)		No	Traffic will be NAT bypassed.

‌

For more information, in the System Administration Guide, see the Session Recovery and Interchassis Session Recovery chapters.

 

NAT64 Overview

Stateful NAT64 is a mechanism for translating IPv6 packets to IPv4 packets and vice-versa. The IPv4 address of IPv4 server/host in an IPv4 network is obtained to and from IPv6 addresses by using the configured stateful prefix. The IPv6 addresses of IPv6 hosts are translated to and from IPv4 addresses by installing mappings in the usual NAT manner. The following figure illustrates the working of NAT64 with DNS64.

 

NAT64 Mechanism

NAT64 is applied on traffic based on the rule match (Destination based NATing). If NAT64 has to be applied, then the NAT64 will translate and forward them as IPv4 packets through the IPv4 network to the IPv4 receiver. The reverse takes place for packets generated by hosts connected to the IPv4 network for an IPv6 receiver. If NAT64 is not applied on the IPv6 packet, then the IPv6 packet will not be translated and sent as is (NAT bypassed) and will be routed within the IPv6 network to the destination.

NAT64 will not be applied for packets whose destination IP address does not match a pre-defined prefix. NAT64 will be applied only for packets whose destination IP address matches a pre-defined prefix. The pre-defined prefix is configurable and it is a single prefix.

 

NAT64 Translation

For NAT64, Network address translation and Protocol translation are done on the packets. The uplink IPv6 packets that are destined to hosts in the IPv4 network must be protocol translated to IPv4 packets and forwarded. The downlink IPv4 packets destined to hosts in IPv6 network must be protocol translated to IPv6 packets and then forwarded.

The Network address translation is done using the following ways:

•	One-to-One NAT: In the case of 1:1 NAT, the subscriber IPv6 address is uniquely mapped to a given NAT IPv4 address. Port translation is not done as the NAT IP address is associated with a single subscriber and not shared by many users.

•

Many-to-One NAT: In the case of N:1 NAT, the subscriber IPv6 address and source port is mapped to a given NAT IPv4 address and NAT port. Port translation must be done as the same NAT IPv4 address is shared by multiple users. Hence, the L4 ports must be translated to differentiate the connections originating from multiple users sharing the same NAT IPv4 address.

 

Limitations for One-to-One NAT64

This section lists the limitations for One-to-One NAT64 in this release.

•	One-to-One NAT IP allocated to a subscriber can either be used for NATing IPv4 traffic or IPv6 traffic from a given subscriber but not both simultaneously.

•	In the case of One-to-One NAT, a given destination can be associated with only one prefix at any point of time as maintained in the destination prefix list. If the same destination has to be associated with multiple prefixes, then such packets will be dropped.

•	Any downlink traffic received on One-to-One NAT IP will always be translated to the same 128-bit IPv6 address (though interface IDs can actually be different).

•	One-to-One NAT IP status is lost after recovery. The NAT IP that was previously used for NAT44 or NAT64 is not recovered. Based on the first packet that is received after call recovery and the PDN type, the IP will be used for NATing IPv4 or IPv6 traffic.

 

Protocol Translation

This section describes the Uplink and Downlink Packet translation.

•

Uplink Packet Translation: The uplink packets are translated from IPv6 to IPv4. The IP headers in the packet will be translated. The existing NAT APIs are enhanced to perform Protocol translation. Along with the NAT mapping, the prefix/suffix to be used for translation will also be passed. In case of fragmented packets, the packets need to be reassembled and then translated. The uplink packet translation includes:

•	IPv6 to IPv4 Header Translation: The original IPv6 header on the packet is removed and replaced by an IPv4 header.

•	ICMPv6 to ICMPv4 Header Translation: The original ICMPv6 header on the packet is removed and replaced by an ICMPv4 header.

•	Packet Translation

•

Downlink Packet Translation: The downlink packets need to be translated from IPv4 to IPv6. The existing NAT APIs are to be enhanced to perform Protocol translation. Along with the NAT mapping, the prefix/suffix to be used for translation will also be passed. In case of fragmented packets, the packets need to be reassembled and then translated. The downlink packet translation includes:

•	IPv4 to IPv6 Header Translation: The original IPv4 header on the packet is removed and replaced by an IPv6 header.

•	ICMPv4 to ICMPv6 Header Translation: The original ICMPv4 header on the packet is removed and replaced by an ICMPv6 header.

 

NAT64 ALGs support

NAT64 ALGs support the following protocols:

•	File Transfer Protocol (FTP)

•	Point-to-Point Tunneling Protocol (PPTP)

•	Real Time Streaming Protocol (RTSP)

•	Trivial File Transfer Protocol (TFTP)

 

How NAT Works

The following steps describe how NAT works:

Step 1

In the subscriber profile received from the AAA Manager, the SessMgr checks for the following:

•	Enhanced Charging Service subsystem must be enabled

•	In the Firewall-and-NAT policy, NAT must be enabled

•	The Firewall-and-NAT policy must be valid

•	For Many-to-One NAT, at least one valid NAT IP pool must be configured in the Firewall-and-NAT policy, and that NAT IP pool must be configured in the context

Step 2

If all of the above is true, once a private IP address is allocated to the subscriber, the NAT resource to be used for the subscriber is determined. This is only applicable for not-on-demand allocation mode.

Important: The private IP addresses assigned to subscribers must be from the following ranges for them to get translated: Class A 10.0.0.0 – 10.255.255.255, Class B 172.16.0.0 – 172.31.255.255, and Class C 192.168.0.0 – 192.168.255.255

Important: A subscriber can be allocated only one NAT IP address per NAT IP pool/NAT IP pool group from a maximum of three pools/pool groups. Hence, at any point, there can be a maximum of three NAT IP addresses allocated to a subscriber.

Step 3

Flow setup is based on the NAT mapping configured for the subscriber:

•	In case of one-to-one NAT mapping, the subscriber IP address is mapped to a public IP address. The private source ports do not change. The SessMgr installs a flow using the NAT IP address and a fixed port range (1–65535).

•

In case of many-to-one NAT mapping, a NAT IP address and a port from a port-chunk, are allocated for each connection originating from the subscriber. In order to identify a particular subscriber call line, the SessMgr installs a flow using NAT (public) IP address + NAT ports allocated for the subscriber.

The following figures illustrate the flow of packets in NAT processing.

 

NAT Processing Flow

 

... NAT Processing Flow

 

... NAT Processing Flow

 

... NAT Processing Flow

 

 

Cisco Systems Inc.

Tel: 408-526-4000

Fax: 408-527-0883