This chapter covers the following subjects:
• Describe the operation of a Cisco IOS Software GET VPN: Covers a new category of VPNs that does not require a tunnel GET VPN and provides large-scale, connectionless, tunnel-free transmission protection that takes advantage of an existing routing infrastructure.
• Plan the deployment of Cisco IOS Software GET VPN: Covers gathering input information and making deployment decisions about your environment.
• Configure and verify a Cisco IOS Software GET VPN key server: Covers the functionality of the GET VPN key server and its configuration.
• Configure and verify Cisco IOS Software GET VPN group member: Covers the deployment of GET VPN group members.
• Configure and verify high-availability mechanisms in a GET VPN: Covers configuring high-availability mechanisms in a GET VPN topology.
The Cisco Group Encrypted Transport Virtual Private Network (GET VPN) technology is a solution that allows easy deployment of a complex, redundant, fully meshed VPN.
The “Do I Know This Already?” quiz helps you decide whether you really need to read the entire chapter. If you already intend to read the entire chapter, you do not necessarily need to answer these questions now.
The eight-question quiz, derived from the major sections in the “Foundation Topics” portion of this chapter, helps you determine how to spend your limited study time.
Table 19-1 outlines the major topics discussed in this chapter and the “Do I Know This Already?” quiz questions that correspond to those topics.
Table 19-1. “Do I Know This Already?” Foundation Topics Section-to-Question Mapping

The goal of self-assessment is to gauge your mastery of the topics in this chapter. If you do not know the answer to a question or are only partially sure of the answer, you should mark this question wrong for purposes of the self-assessment. Giving yourself credit for an answer that you correctly guess skews your self-assessment results and might provide you with a false sense of security.
1. GET VPNs use which feature to provide large-scale transmission protection that uses the existing routing infrastructure? (Select all that apply.)
a. Tunnel-free
b. X.500
c. Connectionless
d. ISAKMP
e. Encrypted
2. GET VPNs use a concept of which of the following to provide transmission protection? (Select all that apply.)
a. Certificates
b. IPsec
c. Key servers
d. Group members
e. None of these answers are correct.
3. To implement a GET VPN over the Internet, which type of IP addresses must be used on all networks?
a. Private
b. Class A
c. NAT
d. Routable
e. None of these answers are correct.
4. GET VPNs maintain which aspect of the data packet?
a. Original IP header
b. Size
c. MAC address
d. Don’t Fragment bit setting
e. None of these answers are correct.
5. Which of the following are the two choices of rekeying used by key servers?
a. Unicast
b. Symmetric
c. Asymmetric
d. Multicast
6. Which of the following do you configure to prevent traffic from traversing an untrusted interface unless the group member is registered into a GET VPN?
a. ACL
b. Policy map
c. Fail-closed policy
d. GET VPN key server
e. None of these answers are correct.
7. What event might lead to several independent groups of key servers rekeying group members with different session keys?
a. Network split
b. Route reconvergence
c. Network merge
d. None of these answers are correct.
The answers to the “Do I Know This Already?” quiz are found in Appendix A. The suggested choices for your next step are as follows:
• 6 or less overall score: Read the entire chapter. This includes the “Foundation Topics” and the “Exam Preparation” sections.
• 7 or 8 overall score: If you want more review on these topics, skip to the “Exam Preparation” section. Otherwise, move on to Chapter 20, “Deploying Remote Access Solutions Using SSL VPNs.”
Fully meshed VPNs are typically challenging from a scalability and manageability standpoint. These kinds of deployments with a large number of sites are typically avoided because of their complexity. This chapter will cover the fundamentals of the Cisco IOS Software GET VPN technology, how to plan its deployment, how to configure and verify GET VPN key servers and members, and how to implement high availability for GET VPNs.
Beginning with Cisco IOS Software Release 12.4(11)T, GET VPNs provide large-scale, connectionless, tunnel-free transmission protection that takes advantage of an existing routing infrastructure. Even though it can be used with Multiprotocol Label Switching (MPLS), IP, Frame Relay, and ATM networks, it is an ideal cryptographic solution for MPLS VPNs that need site-to-site encryption.
With the introduction of GET VPNs, Cisco makes available a new category of VPN that does not require a tunnel. By removing the need to establish point-to-point tunnels, branch networks can scale larger and still maintain the network intelligence features that are necessary for voice and video quality, such as quality of service (QoS), routing, and multicast. GET VPNs provide a new standards-based IPsec security model that uses the concept of “trusted” group members. Trusted member routers use a common security methodology that is independent of requiring any point-to-point IPsec tunnel relationship.
GET VPNs can be deployed in a variety of WAN topologies, including IP and MPLS. MPLS VPNs that use GET achieve high availability, manageability, and cost-effectiveness while meeting transmission protection requirements. GET VPNs provide a flexibility that enables enterprises to manage their own security over a service provider WAN or to offload the encryption services to the provider. GET VPNs simplify large Layer 2 or MPLS networks that need partial or full meshed connectivity.
When deploying IPsec, it is recommended that hardware acceleration for IPsec be included in the chosen router platform to avoid performance degradation.
IPsec acceleration and the GET feature set are supported on the onboard cryptographic capabilities of the Cisco Integrated Services Routers (ISR), the Cisco 7200 Series routers, and the Cisco 7301 router with VPN modules. Refer to Cisco.com for support offered by other Cisco platforms.
Group controller/key servers (GCKS), also known as key servers (KS), and group members are the two key components that comprise the GET VPN architecture. The key server authenticates all group members, performs admission control to the GET VPN domain, and creates and supplies group authentication key as security associations (SA) to group members. Group members provide transmission protection to sensitive site-to-site (member-to-member) traffic. Figure 19-1 illustrates the relationship between a key server and group members.
Figure 19-1. GET VPN Key Server and Group Member Routers

Key servers distribute keys and policies to all registered and authenticated group member routers. Key distribution and management are made easier because of the centralized distribution of keys and policies.
All communication between a key server and group members is encrypted and secured using the Internet Key Exchange (IKE) Group Domain of Interpretation (GDOI) protocol. IKE GDOI is a standards-based Internet Security Association and Key Management Protocol (ISAKMP) group key management protocol that provides secure group communications. GET VPNs use IKE GDOI as the group keying mechanism.

IKE GDOI supports the use of two keys: Traffic Encrypting Key (TEK) and Key Encrypting Key (KEK):
• TEK: A key that is used to protect traffic between group members
• KEK: A key this is used to protect rekeys (during a key refresh) between key servers and group members
The TEK is distributed to all group members by the key server, and they use the TEK to communicate to members of the group and to create and verify IPsec packets. The KEK is also distributed to group members who in turn use it to decrypt incoming rekey messages from the key server.
When a registration message is received, the key server generates information that contains the rekey policy (one KEK) and the new IPsec SAs (multiple TEK attributes, traffic encryption policy, lifetime, source and destination information about the traffic that needs to be protected, and the security parameter index (SPI)-ID that is associated with each TEK). The newly created IPsec SAs are then sent to the group members. The key server maintains a table that contains the IP address of each group member and its group association. When a group member registers, the key server adds the new IP address to its associated group table.
Key servers and group members can simultaneously support multiple groups.
GET VPN group members that have valid group IPsec SAs assume that traffic they encrypt can be decrypted by some other legitimate GET VPN group member.
In a traditional multicast implementation, the sender does not know who potential recipients are. Using Protocol Independent Multicast - Sparse-Mode (PIM-SM), the multicast router sends multicast traffic to the configured rendezvous point (RP). The RP maintains a list of the multicast group recipients.
With GET VPNs, the sender assumes that the legitimate group members obtain a TEK from the group key server. The group member encrypts the multicast data, with header preservation, and the packet is switched out of the router. The replication of the multicast packet is performed in the core based on the source and multicast group IP address that is retained in the multicast data packet.
The secure data plane unicast packet is similar to the multicast example. In the secure data plane, the receiver does not know who potential encryption sources might be. The receiver assumes that the legitimate group members obtained a TEK from the group key server. The receiver authenticates the group membership when it is able to decrypt the data packet.
GET VPNs provide the same security benefits that IPsec provides. These security parameters include packet confidentiality (using cryptographic encryption algorithms such as AES [Advanced Encryption Standard] or 3DES [Triple Data Encryption Standard]), packet integrity and authenticity (using cryptographic HMAC [Hash Message Authentication Code] algorithms such as SHA-1 [Secure Hash Algorithm-1] HMAC or MD5 [Message Digest 5] HMAC), and a Cisco-proprietary time-based (instead of sequence number–based) antireplay protection mechanism. The standard sequence number method cannot be used in VPNs because synchronizing sequence number counters across many members is not feasible.
AES encryption is the recommended algorithm to use in GET VPNs. Because of the possibility of having large numbers of VPN group members using the same (group) session keys, and a fairly short initialization vector (IV) of 3DES, the use of 3DES in GET VPNs is not recommended.
Note that Network Time Protocol (NTP) is not required in GET VPNs (unless certificate-based authentication is used), because group members use proprietary pseudotime instead of standard time to create and verify timestamps.
The GDOI, Group Domain of Interpretation, is the underlying standard for GET VPN as defined in RFC 3547. GDOI defines a key management protocol that is based on IKE/ISAKMP. It uses the same principles as IKE/ISAKMP to generate symmetric encryption keys but uses two keys, KEK and TEK. This key management protocol is an extension of IKE/ISAKMP and uses UDP port 848. One significant difference is that GDOI IKE SAs do not need to linger between members after initial establishment, but they can be quickly expired after a group member has authenticated to the key server and obtained the group policy. The second major difference is that GDOI IKE sessions do not get established between all peers in a VPN, only between each group member and the key server (or multiple key servers for redundancy).
One other notable difference is that all group members use the same set of session keys to protect network traffic. This is more efficient when compared to traditional IPsec VPNs in which each pair of peers has a private set of IPsec SAs that are shared only between those two peers.
GET VPNs use rekey messages to refresh their IPsec SAs (session keys) outside of IKE sessions. When the group IPsec SAs are about to expire, one single rekey message for a particular group is generated on the key server. Distribution of the rekey message does not require that new IKE sessions be created. GET supports rekeying for unicast and multicast.
If any part of the enterprise network is not multicast capable, the unicast transport mechanism should be used to distribute the rekey message for all group members. The key server will send a separate rekey for every group member, and the group member must respond to the key server with an acknowledgment message. The key server will retransmit rekeys if it fails to receive the acknowledgment from the group member. If it fails to get a response from the group member after three rekeys, it removes the group member.
The key server maintains a list of the registered group members in the database and tracks the number of rekeys that have been sent and the acknowledgments received per group member. This database of information is extremely helpful in troubleshooting issues with a specific group member.
If the enterprise network is multicast capable, using multicast rekeying is recommended because it is more scalable.

The following are some general guidelines to consider regarding rekeying:
• If most of the group members are only unicast capable, use unicast rekeying.
• If most of the group members are capable of multicast and the entire enterprise network is capable of multicast, use multicast rekeying.
Table 19-2 provides some information that should be considered when deciding between using unicast or multicast rekeying.

Table 19-2. Unicast Versus Multicast Rekeying Methods

When the key server sends out multicast rekeys to the group members, it sends a single multicast rekey packet to the core and the core replicates the packet for each multicast group member. Because multicast rekeying does not require acknowledgment, the rekeys will be retransmitted two or three times during the rekey period. Multicast rekeying for very large environments is more efficient and scalable because it makes use of the multicast replication provided by the core; therefore, it is the recommended rekeying method. Multicast rekeying also reduces key server load by drastically reducing the number of rekeying messages it must process.
To use multicast transport for rekeying, the entire network must be multicastcapable, including the MPLS/IP core. That means Multicast VPN (MVPN) is required on the MPLS core.
When the key server uses unicast rekeying, the key server generates rekey messages for only a few group members at a time and makes sure that all group members receive the rekey message for the new SA before the old SA expires. This helps reduce the possibility of latency issues. Also, when the group members receive the rekey message from the key server, the group member sends an encrypted acknowledgment to the key server using the key received as part of the rekey message. This process keeps the list of active group members up to date and ensures that rekey messages are sent only to active group members.
In a large network, unicast rekeying creates a large load on the key server because it must send unicast rekey messages and process all the acknowledgments received from every group member.
The number of retransmit attempts and the retransmit interval are configurable. As stated previously, the key server will remove a group member from which the key server fails to receive acknowledgment three times. The group member must fully reregister with the key server after its current IPsec SA expires to receive rekey messages. If the group member does not receive a rekey before the TEK expires, it reregisters with the key server before the current IPsec SA expires.
In a large network, if there is a small part of the network that does not support multicast, the multicast transport mechanism can still be used for rekeying. This method of rekeying causes the small set of unicast group members to reregister, but it places a lower load on the key server than if the key server had to use the unicast transport with every group member. Each unicast member that is reregistering will do so before the current group key expires, which avoids any loss of data.
It is imperative for group members to synchronize the removal of old SAs and the installation of new SAs. As new SAs are received through rekey messages, outgoing data packets are still encrypted using the old SAs, but incoming packets are decrypted using both the old and the new SAs for a period of time. After a specified amount of time (T1), outgoing packets are encrypted using new SAs while incoming packets are still decrypted using both old and new SAs for a period of time. At the next interval (T2), old SAs are removed and new SAs encrypt and decrypt the traffic. T1 and T2 are set internally to 30 seconds.
The benefit of GET VPNs is that they use the existing routing infrastructure rather than adding a tunnel-based overlay like traditional IPsec. Figure 19-2 shows how GET VPN data packets maintain their original IP source and destination addresses. This enables organizations to use existing Layer 3 routing information, which can address multicast replication inefficiencies and improve network performance. Encrypting multicast packets with IP address preservation is necessary to preserve the (S,G) (source, multicast group) information so that replication of multicast packets in the cores can be based on the original (S,G) information.

Figure 19-2. GET Packet Structure

GET VPNs have the following benefits:
• Very scalable in that the configuration does not grow significantly when adding group members in a full mesh
• Provides scalable support for multicast traffic
GET VPNs also have the following limitations:
• VPN addresses must be routable in the transport network. This is because of the use of the original IP header, and in most cases, it prevents GET VPNs from being used over the Internet.
• The compromise of one peer has a larger effect because the group shares session keys.
• Key servers must be available during rekeys and registration for the entire network to operate.
A successful GET VPN deployment requires that you gather information about your environment and then make deployment decisions about your environment. The following sections cover these activities.

The following input parameters need to be determined and analyzed when deploying GET VPNs:
• If using existing devices, ensure that hardware and software on the routers can adequately handle the cryptographic performance that is required and that the current software release supports GET VPNs on the existing platform.
• Determine security requirements to deploy the correct data protection (algorithms, key lengths, and key lifetimes).
• Carefully define traffic flows that require protection by the GET VPNs. Some traffic flows will need to be forwarded in the clear.
• Determine whether high availability is needed, and ensure that the proper transport network and redundant equipment exist.

The basic deployment tasks for implementing GET VPNs are as follows:
Task 1: Configure IKE sessions between each GET member and key server.
Task 2: Configure an IPsec traffic protection policy on the GET VPN key server.
Task 3: Configure GET VPN members to register into a GET VPN group.
Task 4: Optionally, tune local policies on members to override policies received from the key server.
GET VPN deployments require that decisions be made based on network requirements and capabilities. For IKE peer authentication, choose to use either pre-shared keys (PSK) or public key infrastructure (PKI). Pre-shared keys are easier to deploy, but pre-shared key–based main mode in IPsec does not provide support for dynamically addressed group members. High availability can be achieved by using multiple key servers. This simply adds configuration overhead. Local policies should be tuned when using dynamic routing protocols or in-band management protocols.

Three main guidelines to keep in mind when implementing a GET VPN are as follows:
• Use GET VPNs for scalable, site-to-site, full-mesh connectivity for a large number of sites.
• To implement GET VPNs over the Internet, routable IP addresses on all networks must be used.
• There is no scalability issue with deploying using pre-shared keys because only a limited number of IKE sessions are required (each group member to each key server). Other criteria, such as security requirements, must be taken into account when deciding between PSK and PKI.
The GET VPN key server is a critical part of the GET VPN architecture. Most of the GET VPN configuration is done on the key server. Some of the configuration items are
• Group member authentication and admission control
• Traffic protection policy (definition of IPsec SA parameters)
• Rekeying policy (unicast or multicast, key lifetimes, retransmissions)
These configuration items are used by all group members unless they are overridden by a local policy configured on a specific group member.

Perform the following configuration sequence to configure a GET VPN key server:
Task 1: (Optional) Configure an IKE policy. The default IKE policy can be used.
Task 2: Generate/configure authentication credentials for all group members.
Task 3: Generate or choose existing RSA keys on the key server to authenticate rekeys.
Task 4: Configure a traffic protection policy.
Task 5: Enable and configure the GET VPN key server function itself.
Task 6: (Optional) Tune the rekeying policy.
Task 7: Create a GET VPN crypto map and apply it on a key server interface. This enables the key server to listen for group member registration requests and distribute the configured traffic protection policy.
Two of the configuration tasks are optional:
• Configure a nondefault IKE policy. You might want to configure an IKE policy that has higher security settings than those in the default policy.
• Tune the rekeying policy to use unicast rekeying because multicast is the default. This is required if the transport network does not support multicast.
Figure 19-3 illustrates the configuration scenario used for the configuration sequence that follows.
Figure 19-3. GET VPN Configuration of Key Server

The configuration sequence involves a full key server router configuration with the following policy:
• IKE policy that uses pre-shared key authentication of peers and uses DH group 14 for the initial key exchange. All other IKE parameters will be left at their defaults.
• The key server traffic protection policy should require protection for all traffic with source or destination addresses of 10.0.0.0/8 and also require AES 128-bit encryption and SHA-1 HMAC algorithms.
• The GET VPN should use unicast rekeying that is authenticated with an RSA key pair that should be generated on the key server for that explicit purpose.
Task 1 is to optionally create an IKE policy with the crypto isakmp policy global configuration command. Example 19-1 shows an IKE policy specifying RSA signatures for authentication and Diffie-Hellman group 14 to protect the key exchange.

Example 19-1. Configure an Optional IKE Policy
Router(config)# crypto isakmp policy 10
Router(config-isakmp)# authentication pre-share
Router(config-isakmp)# group 14
IKE peering takes place only between each group member and the key server.
There must be an IKE policy on each group member that matches that of the key server.
The second task of configuring the key server involves configuring authentication credentials. Pre-shared keys will be used as the authentication method. IKE main mode uses a unique PSK for each peer that is tied to a specific IP address. Example 19-2 shows the commands.

Example 19-2. Generate and Configure Authentication Credentials
Router(config)# crypto isakmp key ad73asmdkfl902380amadfjkasdjf 172.17.2.4
Router(config)# crypto isakmp key akjsdfljfdasdfu2389872jh3241u 172.17.0.1
Task 3 creates the keys that will be used for rekeying authentication. To use redundant key servers, the generated keys must be exportable so that they can be imported on all other key servers. Example 19-3 shows a key being generated. Note that GET VPNs must have the same rekey RSA key pair on all key servers.

Example 19-3. Generate RSA Keys
Router(config)# crypto key generate rsa modulus 2048 label MYRSAKEYS exportable
Task 4 involves the configuration of the IPsec profile, an IPsec transform set, and a traffic flow specification.
The IPsec profile defines the encapsulation and cryptographic settings that will be distributed to the group members by the key server as part of the SA. The IPsec profile usually defines the transform set, session key lifetimes, and the requirements for Perfect Forward Secrecy (PFS).
The IPsec transform set defines a combination of security protocols, algorithms, and other settings to apply to IPsec-protected traffic. During the SA negotiation, peers agree to use a particular transform set for certain data flows. The transform set specifies one of two IPsec security protocols (Authentication Header [AH], Encapsulating Security Payload [ESP], or both) and specifies which algorithms to use with the chosen security protocol. Example 19-4 shows the configuration of the IPsec profile and transform set parameters.
Also in Task 4, traffic that is to be protected by the GET VPN IPsec profile must be defined. To do so, define an access control list (ACL) on the key server to define the traffic that must be protected. This information is pushed to the group members by the key server during their registration process. This allows the key server to choose or change the group policy dynamically as needed. The ACL can use both deny and permit ACL lines on the key server, which allows more granular control over the traffic that must be cryptographically protected. A deny entry in the ACL signifies traffic that is to be forwarded in the clear.
An ACL that is configured on the group member takes precedence over what is downloaded from the key server. The local ACL entries will be used at the top of the GET VPN ACL and then followed by the entries that are downloaded for the key server.

Example 19-4. Configure Traffic Protection Policy on the Key Server
Router(config)# crypto ipsec tranform-set MYSET esp-aes esp-sha-hmac
Router(config)# crypto ipsec profile MYIPsecPROFILE
Set transform-set MYSET
Router(config)# ip access-list extended MYGETVPNACL
Router(config-acl)# permit ip 10.0.0.0 0.255.255.255 10.0.0.0 0.255.255.255

Task 5 defines key server parameters and creates a policy that group members will use upon registration. Initially, use the crypto gdoi command to create a crypto GDOI group (a logical GET VPN). Use the identity command in GDOI group configuration mode to set the identity of the group to either an IP address or a number. The identity distinguishes the specific group configuration, because there can be multiple GET VPN groups on each key server or member.
The GDOI server group configuration command identifies that the group key server is defined locally on this router. Example 19-5 shows the configuration example.
The IPsec profile defines the encapsulation and cryptographic settings that will be distributed to the group members by the key server as part of the SA. The IPsec profile usually defines the transform set, session key lifetimes, and the requirements for Perfect Forward Secrecy (PFS).
Example 19-5. Enable the GET VPN Key Server Function
Router(config)# crypto gdoi group MYGETVPNGROUP
Router(config-gdoi)# identity number 1324
Router(config-gdoi)# server local
Router(config-gdoi)# address ipv4 172.17.0.1
Router(config-gdoi)# sa ipsec 10
Router(config-gdoi-sa)# profile MYIPsecPROFILE
Router(config-gdoi-sa)# match address ipv4 MYGETVPNACL
Task 6 of key server configuration involves tuning the rekey options. Options that are available are rekey authentication, rekey transmit option, and a source address for rekeys. Configuring rekey options as part of the registration policy on the key server replaces the need to do it on all group member routers. Multicast is the default transport type for rekey messages.

The following options cover when to use multicast or unicast rekey transport:
• If all members of a group are multicast capable, use no rekey transport unicast (this means that multicast rekeying will be used).
• If all the members of a group are only unicast capable, use rekey transport unicast.
• If there is a mix of members in a group, and the majority of the members are multicast capable and only a few are unicast capable, use no rekey transport unicast.
In the last case, all unicast members must reregister because they do not receive the multicast rekeys, but this is less of a load on the key servers. Refer to Example 19-6 for the configuration commands.

Example 19-6. Tuning the Rekey Policy
Router(config-gdoi)# rekey transport unicast
Router(config-gdoi)# rekey authentication mypubkey rsa MYRSAKEYS
The last task of key server configuration creates a GDOI crypto map and applies that map to the interface. The crypto map entry must be of the type gdoi and reference the GDOI group created in Task 5 in this sequence. The crypto map must be applied to the interface to which group members will register. Example 19-7 shows the commands.
Example 19-7. Create and Apply the GET VPN Crypto Map
Router(config)# crypto map MYCRYPTOMAP 10 gdoi
Router(config)# set group MYGETVPNGROUP
Router(config)# interface GigabitEthernet0/0
Router(config-if)# crypto map MYCRYPTOMAP
Cisco Configuration Professional 2.0 supports the creation of GET VPN key servers on a managed device. It allows the key server to be configured either through a configuration wizard or by manually entering all the options.
The show crypto gdoi command displays information about the local GET VPN configuration. As demonstrated in Example 19-8, the output shows the GET VPN key server settings and the number of registered group members.
Example 19-8. show crypto gdoi Command
Router# show crypto gdoi
GROUP INFORMATION
Group Name : MYGETVPNGROUP (unicast)
Group Identity : 1234
Group Members : 34
IPSec SA Direction : Both
Group Rekey Lifetime : 86400 secs
Rekey Retransmit Period : 10 secs
Rekey Retransmit Attempts : 2
IPSec SA Number : 8
IPSec SA Rekey Lifetime : 3600 secs
Profile Name : MYIPSECPROFILE
Replay Method : Count Based
Replay Window Size : 64
ACL Configured : access-list MYGETACL
Group Server List : Local
The show crypto gdoi ks rekey command displays detailed rekey settings that are active for a GET VPN group, as demonstrated in Example 19-9.
Example 19-9. show crypto gdoi ks rekey Command
Router# show crypto gdoi ks rekey
Group MYGETVPNGROUP (Unicast)
Number of rekeys sent : 23843
Number of rekeys retransmitted : 56
KEK rekey lifetime (sec) : 86400
Number of retransmissions : 2
IPSec SA 10 lifetime (sec) : 3600
On the key server router, the show crypto gdoi ks members command displays information about registered group members, as demonstrated in Example 19-10.
Example 19-10. show crypto gdoi ks members Command
Router# show crypto gdoi ks members
Group Member Information:
Number of rekeys sent for group getvpn : 10
Group member ID : 172.17.2.24
Group ID : 1234
Group Name : MYGETVPNGROUP

Consider the following implementation guidelines when configuring a GET VPN key server:
• Configure IKE peering between each group member and the key servers as in a traditional IPsec VPN, using pre-shared keys or PKI for authentication.
• Identify traffic that does not require IPsec protection and use deny statements on the ACL on the key server to exclude that traffic from the GET VPN protection.
• On the key server, make the ACL length as short as possible to minimize the number of SAs on the peers. Also, ensure that the ACL is symmetric and matches flows in both directions.
• If interface redundancy is required, bind the key server to a loopback interface. Ensure that the loopback is included in the crypto map definition.
GET VPN group members require very little effort because GET VPN key servers distribute most of the VPN settings to the group members. On the group member, the following must be configured:
• Key server IP address: Configure a list of all available key servers for a GET VPN group.
• Key server IKE authentication details: Specify the key server authentication credentials (pre-shared keys or PKI).
• Local policy overrides: Optionally, configure local policy overrides that provide exceptions to the policy that is downloaded from the key server.

Perform the following configuration sequence to configure a GET VPN group member:
Task 1: (Optional) Configure an IKE policy. The default IKE policy can be used.
Task 2: Generate/configure authentication credentials for all group members. These credentials must match the IPsec configuration on the key server.
Task 3: Enable and configure the GET VPN group member function.
Task 4: Create a GET VPN crypto map, and apply it on an untrusted group member network interface.
Task 5: (Optional) Configure a fail-closed policy. This is dependent upon the local security policy.
Two of the configuration tasks are optional:
• Configure a nondefault IKE policy: You might want to configure an IKE policy that has higher security settings than those in the default policy. Also, consider configuring shorter SA lifetimes to reduce the load on the key server because IKE SAs do not need to remain after registration and policy download.
• Consider using a fail-closed policy: This enhances the security posture and prevent packets from traversing the untrusted network while IPsec SAs are not yet established. There should still be exceptions to this to allow routing protocols and management traffic.
Figure 19-4 illustrates the configuration scenario used for the configuration sequence that follows.
Figure 19-4. GET VPN Configuration of Group Member Routers

The configuration sequence involves configuring a GET VPN group member with the following policy parameters:
• IKE policy that uses pre-shared key authentication of peers and uses DH group 14 for the initial key exchange. All other IKE parameters will be left at their defaults.
• The untrusted network interface will not permit any user traffic through unless the group member has successfully registered in to the GET VPN.
• The fail-closed policy should permit Secure Shell (SSH) to all group member loopback IP addresses (10.x.255.1 where x is unique for each member).
Task 1 is to optionally create an IKE policy with the crypto isakmp policy global configuration command. Example 19-8 shows an IKE policy specifying pre-shared keys for authentication, Diffie-Hellman group 14 to protect the key exchange, and setting the IKE lifetime to 300 seconds.

Example 19-11. Configure an Optional IKE Policy and Authentication Credentials
Router(config)# crypto isakmp policy 10
Router(config-isakmp)# authentication pre-share
Router(config-isakmp)# group 14
The IKE configuration on the group member must match that of the key server, except for the IKE lifetime, which will negotiate down to the shortest configured time.
There must be an IKE policy on each group member that matches that of the key server.
The second task of configuring the key server involves configuring authentication credentials. Pre-shared keys (PSK) will be used as the authentication method. Example 19-12 shows the commands.

Example 19-12. Generate and Configure Authentication Credentials
Router(config)# crypto isakmp key ad73asmdkfl902380amadfjkasdjf 172.17.2.4
Router(config)# crypto isakmp key akjsdfljfdasdfu2389872jh3241u 172.17.0.1
Task 3 enables and configures the GET VPN group member function on the router. The subtasks that follow are needed and configured in Example 19-13:
• Create a GDOI group and assign a locally significant name. Use the crypto gdoi group command.
• Set up the group identity number to match the identity number configured on the key server.
• Provide the IP address of the key server.

Example 19-13. Enable and Configure GET VPN Group Member
Router(config)# crypto gdoi group MYGETVPNGROUP
Router(config-gdoi)# identity number 1324
Router(config-gdoi)# server address ipv4 172.17.0.1
Task 4 creates the GET VPN crypto map and then applies it to the untrusted network interface on the group member router.
The set group command in crypto map configuration mode maps the configured GDOI group to the crypto map. Finally, the crypto map interface configuration command applies the crypto map to the interface. See Example 19-14 for the configuration usage.

Example 19-14. Create and Apply Crypto Map
Router(config)# crypto map MYCRYPTOMAP 10 gdoi
Router(config-map)# set group MYGETVPNGROUP
Router(config)# interface GigabitEthernet0/0
Router(config-if)# crypto map MYCRYPTOMAP
Task 5, the last and optional task of configuring a GET VPN group member router, is completely dependent upon an organization’s security policy. Configuring a fail-closed policy means that no traffic can traverse through the untrusted network interface unless the member has successfully registered with the key server. Adding the fail-close option to the crypto map MYCRYPTOMAP gdoi command along with the activate command in crypto map configuration mode will enable the fail-closed feature. You should configure an ACL that excludes management traffic such as SSH. Deny statements in the ACL will allow specified traffic to traverse the interface in clear text. After the ACL is defined, use the match address command in crypto map configuration mode. Example 19-15 shows the configuration example.
Example 19-15. Configure a Fail-Closed Policy
Router(config)# crypto map MYCRYPTOMAP gdoi fail-close
Router(config-map)# match address MYFAILCLOSEACL
Router(config-map)# activate
As with the configuration of the GET VPN key server, Cisco Configuration Professional 2.0 supports the creation of GET VPN group members and can also be done either through a configuration wizard or by manually entering all the options.
The show crypto gdoi command displays information about the status of the group member’s registration. It also displays the key server that is defined in the configuration.

Consider the following implementation guidelines when configuring a GET VPN group member router:
• Lowering IKE lifetimes on group members is recommended to reduce the load on the key server.
• In most cases, it is highly recommended to use the fail-closed feature. Management traffic and other overhead traffic that do not require IPsec protection should be excluded by using deny statements in an ACL to exclude that traffic.
• Performance on the group member will be approximately 5 percent lower for ciphertext when compared to normal IPsec VPN throughput. The more complex time-based antireplay feature of GET takes longer than the typical IPsec sequence-based antireplay.

Perform the following steps for troubleshooting GET VPN group members if the GET VPN IPsec SA does not establish:
Step 1: If the GET VPN IPsec SA does not establish, first test and verify connectivity for each peer. Use ping and traceroute commands between the key server and the group member to assist with verifying reachability.
Step 2: Ensure that the IKE policies on both peers (key server and group member) match. The debug crypto isakmp command can reveal mismatches in IKE policies. The only parameter in an IKE policy that does not have to match is the IKE lifetime parameter.
Step 3: Verify that the IKE SA is authenticated. Verify that pre-shared keys match if this is the authentication method being used.
Step 4: The IKE SA should be established. The show crypto isakmp sa command can show the status. The state of the SA should be GDOI_IDLE and registration should be successful.
Step 5: Verify that the IPsec SA has established. Several commands, such as show crypto gdoi, debug crypto gdoi events, debug crypto gdoi ks, and others can reveal problems with the key server policy being pushed to the group members.
GET VPNs can have multiple key servers configured. These key servers (called cooperative servers) work together to deliver keys and policies to group members. One of the key servers will be elected as the primary server through the use of a proprietary protocol, Cooperative Key Server Protocol (COOP). The primary key server creates IPsec session keys and pushes them to its peer key servers. The peer key servers then distribute session keys to their registered group members. Each group member registers with one of the key servers, which allows them to exchange encrypted data as part of the same VPN. Note that the primary key server distributes the rekey to all group members.
In some cases, there might be several policies and keys being distributed by several cooperating key servers. There can be up to eight key servers on a network. One of the key servers will be elected the primary key server, and the others will synchronize with the aforementioned keys and policies. Key servers can be placed anywhere in the network, as long as they are reachable by the other key servers and the group member routers.
There are two reasons to have multiple key servers:
• Allows a group member router to register with the key server router that is closest for performance reasons
• Allows key server redundancy
Rekey configuration, defined policies, and antireplay settings must be identical on all key servers.
It is recommended to bring up all cooperative key servers first, allow them to synchronize policies and keys, and then configure group members with primary and secondary key servers for registrations.
Cooperative key servers exchange one-way announcement messages, process and store the data content of protocol messages, and update their view of the group state. These announcement messages can contain the following components to help maintain state information:
• Sender priority (identity of a key server): Describes the priority of the sender, which is configurable. The key server with the highest priority becomes the primary key server. If the priority values are the same, the key server with the highest IP address becomes the primary key server.
• Maintaining the sender role: During the synchronization period, geographically disparate key servers might suffer from network latency and more than one key server will become the primary key server. After the latency is gone, the key servers will find each other, exchange role information, and revert to their proper role.
• Request for a return packet flag: All messages are defined as one-way messages. A key server can request state information from a peer to find out its role or request the current state of the group.
• Group state: This is the current IPsec SA and keys that are current for the group. If messages have not been heard from the primary key server for a certain period of time, the Primary Periodic Timer is activated. This causes a secondary key server to send an announcement to the primary key server requesting a return packet. If the secondary key server does not receive a response, the Dead Primary Timer begins. At this point, a reevaluation role time is set when a key server is moving to the primary key server state.
A possible negative aspect to deploying multiple key servers is when connectivity between key servers is down for a long period of time. If this network split occurs, it can lead to several independent groups of key servers, rekeying group members with different session keys. This will lead to connectivity issues between group members, because they will not be able to decrypt traffic from other GET VPN members. After the connectivity is restored, the GET VPN will automatically merge and unify session keys across all members. It is critical that communication among key servers not be interrupted.

Perform the following configuration sequence to configure redundancy for GET VPNs:
Task 1: Key servers in the cluster must have the same RSA key pair. Export the main RSA key pair and import it to the other key servers.
Task 2: Configure a full mesh of IKE sessions between all key servers, which enables the key servers to mutually authenticate.
Task 3: Configure COOP between the key servers. The COOP runs over the IKE sessions between key servers.
Task 4: Configure the same traffic protection policy on all key servers.
Task 5: Configure all group members with multiple key servers for redundancy.
Figure 19-5 illustrates the configuration scenario used for the configuration sequence that follows.
Figure 19-5. GET VPN Redundant Key Servers

The configuration sequence involves configuring redundancy by creating a cluster between two GET VPN key servers with the following policy parameters:
• The key servers will use pre-shared keys for authentication and will use a lifetime of 86400 seconds for their IKE session. All other IKE parameters will be left at their defaults.
• One key server will have a cluster (COOP) priority of 5, and the other router’s priority will be set to 10.
Task 1 is to export the RSA key pair of the key server that was configured first and then import it to all other key servers. Use the crypto key export command to export the key, and then use the crypto key import command to import the key to all other key servers.
The second task of configuring redundancy is to configure IKE peering between all key servers. The IKE lifetime will be set to 86400 seconds because this is between key servers. Also, because of their criticality, issue the crypto isakmp keepalive command to verify mutual reachability of key servers. Example 19-16 shows the command breakdown.
Example 19-16. Configure IKE Peering Between All Key Servers
Router(config)# crypto isakmp policy 10
Router(config-isakmp)# lifetime 86400
Router(config)# crypto isakmp keepalive 10 periodic
Using the crypto gdoi group global configuration command, and then the server local command to enter local server configuration mode, key server redundancy must be enabled with the redundancy command. This will put you in COOP subconfiguration mode. In this mode, set the local key server priority (higher is better) by using the local priority command. Example 19-17 shows the commands in use.
router(config)# crypto gdoi group MYGETVPNGROUP
router(config-gdoi-group)# server local
router(config-local-server)# redundancy
router(gdoi-coop-ks-config)# local priority 10
router(gdoi-coop-ks-config)# peer address ipv4 172.17.0.2
All key server peers in the COOP cluster must be listed in this subconfiguration mode. Peers are specified with the peer address ipv4 command.
Task 4 involves configuring the same policy on all GET VPN key servers. This means repeating Steps 1–3 on all the remaining GET VPN key servers.
Task 5 specifies multiple key servers on the group member using the server crypto gdoi group configuration command, as shown in Example 19-18.
Example 19-18. Configure GET VPN Key Servers
router(config-gdoi)# server address ipv4 172.17.0.1
router(config-gdoi)# server address ipv4 172.17.0.2
To verify IKE peering, verify connectivity between the peers and make sure that the IKE policies match.
The next troubleshooting process is to verify mutual COOP peering. For the local key server role, check the Local KS Role: line in the output of the show crypto gdoi group group_ID ks coop command, as shown in Example 19-19. It should display Primary for the primary key server or Secondary for all others. For each key server peers, the IKE status: should read Established.
Example 19-19. Verify COOP Peering
router# show crypto gdoi ks coop
Crypto Gdoi Group Name :group1
Group handle: 2147483650, Local Key Server handle: 2147483650
Local Address: 1.1.1.1
Local Priority: 100
Local KS Role: Primary , Local KS Status: Alive
Primary Timers:
Primary Refresh Policy Time: 20
Remaining Time: 17
Antireplay Sequence Number: 313
Peer Sessions:
Session 1:
Server handle: 2147483651
Peer Address: 2.2.2.2
Peer Priority: 99
Peer KS Role: Secondary , Peer KS

Consider the following implementation guidelines when configuring a GET VPN COOP cluster and key server redundancy:
• IKE and IPsec SAs must be manually synchronized on all key servers that are in the COOP cluster.
• Place key servers at strategic positions in the network to gain as much efficiency with group member registrations while maintaining connectivity between key servers.
• By distributing group members across multiple key servers and controlling the order of the key servers in the configurations, some load balancing can be achieved.

Perform the following recommended flow for troubleshooting redundant key servers:
Step 1: If the COOP cluster fails to establish, check the key server IKE mesh using the show crypto isakmp sa command.
Step 2: Verify the key server COOP mesh using the show crypto gdoi ks coop, show logging | include COOP, and debug crypto gdoi coop commands.
As mentioned in the section, “How to Use This Book,” in the Introduction, you have several choices for exam preparation: the exercises here, the memory tables in Appendix C, the final exam preparation chapter, and the exam simulation questions on the CD-ROM. The following questions present a bigger challenge than the exam itself because they use an open-ended question format. By using this more difficult format, you exercise your memory better and prove your conceptual and factual knowledge of this chapter. You can find the answers to these questions in Appendix A, “Answers to the DIKTA Quizzes and Fill in the Blanks Questions.”
Review the most important topics in the chapter, noted with the Key Topics icon in the margin of the page. Table 19-3 lists a reference of these key topics and the page numbers on which each is found.


Print a copy of Appendix C, “Memory Tables” (found on the CD), or at least the section for this chapter, and complete the tables and lists from memory. Appendix D, “Memory Table Answers,” also on the CD, includes completed tables and lists to check your work.
Define the following key terms from this chapter, and check your answers in the Glossary:
Internet Key Exchange (IKE) Group Domain of Interpretation (GDOI) protocol
Group Encrypted Transport VPN (GET VPN) Design and Implementation Guide, www.cisco.com/en/US/prod/collateral/vpndevc/ps6525/ps9370/ps7180/GETVPN_DIG_version_1_0_External.pdf.