Network Topology and Redundancy

This section describes the IP routing infrastructure for the DSC Platforms and provides some examples for IP networking requirements in a customer network.

IP Addressing and Networking on the DSC 8000

The DSC 8000 consists of two Ethernet Switches, two management slots (Management CPUs) and multiple routing slots (Routing CPUs). The configuration of the multiple Ethernet Switch ports is as follows:

• one port for management traffic

• one port for SS7/IP or Diameter traffic

• additional ports may be assigned for special functions such as the Integrated Monitoring Feed (IMF)

Internally, the IP networking of the DSC 8000 consists of two independent subnets, with one Ethernet Switch serving each subnet. Each Ethernet Switch is connected to each system card in the DSC 8000 across a dedicated trace. Since the two Ethernet Switches do not forward traffic between them, the internal subnets are effectively isolated.

Each system card uses different Ethernet devices to access each Ethernet Switch. The routing slots have Ethernet devices that map to internal VLANs and indicate the target port on the Ethernet Switch. With the current infrastructure, routing slots have access to routing ports out of the box and to special ports with some customer configuration; routing slots on a DSC 8000 do not have access to management IP ports on the Ethernet Switch.

The example redundancy model in the following illustration is applied for each logical network used by the routing slots.

Each DSC 8000 Routing CPU has two unique IP addresses on two different subnets that can be used to query the PSX. IP Routing is accomplished based on destination IP only.

The actual Ethernet devices used depend on whether the SS7/IP Routing Network is used to communicate with the PSX, or whether we use on the special networks such as the IMF network. The DSC does not typically connect to the PSX using the management IP Network.

For detailed information about the DSC 8000 Ethernet configuration, see Ethernet Connectivity for the DSC 8000.

IP Addressing and Networking on the DSC 8000

IP Addressing and Networking on the DSC SWe

The DSC SWe consists of two VMs with management and routing functionality, and, optionally, several additional Routing VMs. The DSC SWe VM with the routing function has pkt0, pkt1, pkt2 and pkt3 interfaces that can be configured for SS7/IP and Diameter traffic. Some options are supported for redundancy such as

• using pkt0 and pkt1 like the routing devices p19p1.2 and p19p2.4 on the DSC 8000 routing slots
  
  
• configuring VRRP-like redundancy (bonding) for pkt0 on the host

Hence, a DSC SWe VM may have between one and four  IP addresses for SS7/IP and Diameter traffic. Similar considerations apply to logical interfaces for special functions like IMF.

Each DSC SWe VM with the routing function has between one and four local, unique IP addresses that can be used to query the PSX. IP Routing is currently done based on destination IP only. The actual devices used depend on which logical network is used to access the PSX. The DSC  typically does not connect to the PSX using the management IP Network.

For detailed information about DSC SWe networking, see Configuring DSC SWe Network Information.

DSC to PSX Connections and Diameter+ Registrations

The connections between the DSC and PSX exist between routing slots and PSX slaves. For each local address that the DSC uses to communicate with a PSX slave, the DSC must register using a distinct Gateway Name at the PSX. Hence, a logical connection between a DSC routing slot and a PSX slave consists of at least the following information:

• DSC Gateway Name
  
  
• DSC IP Address
  
  
• PSX IP Address
  
  
• PSX UDP Port

The same Gateway Name can be used to register with different PSX slaves, but multiple connections with a single PSX slave require different Gateway Names.

The PSX configuration contains between one or two Gateway Names for each DSC 8000 routing slot, and one to four Gateway Names for each DSC SWe VM with the routing function. If a new DSC routing slot is added, the corresponding Gateway Names must be added to the PSX configuration before that slot can perform PSX queries.

The following illustration depicts two DSC Routing Slots/VMs with two processes each, accessing three PSXs; associated with each line is a registration using the Gateway Name and IP Address shown.

DSC to PSX Connections and Diameter+ Registrations

DSC with Local IP Redundancy (applicable to the DSC SWe)

For a DSC SWe, redundancy of the Ethernet devices may be provided using a bonded device on the host. In that case the DSC and PSX can have interface redundancy while each using a single IP address.

DSC with Local IP Redundancy

PSX with Non-redundant IP Configuration (applicable to both the DSC 8000 and DSC SWe)

The option described in this section is to configure each PSX with a single Ethernet device (no redundancy), or with a bonded Ethernet device (for interface redundancy) and connect each DSC to different PSXs using different local devices on the DSC. This approach is consistent with how the PSX is usually deployed and with current DSC IP routing.

In this case, the failure of Network 1 results in the inaccessibility of PSX1 to the DSC process; all PSX queries would go to PSX 2. More generally, the failure of a DSC Ethernet Switch card would result in the unavailability of half the PSX slaves.

PSX with Non-redundant IP Configuration

PSX Multihoming Configuration (mostly applicable to the DSC 8000)

The PSX can be configured with multiple IP addresses (at the Linux prompt) and the DSC can route to each PSX IP address using different local interfaces. The extra PSX IP would be strictly for the DSCs benefit, so the DSC can do IP routing to different addresses using different interfaces.

In the multihomed configuration depicted in the following illustration, the DSC 8000 slot registers two Gateway Names with the PSX: one associated with IP Address D1_1 and one with D1_2.

UDP datagrams from the DSC to the PSX have source address D1_1 and destination address P1_1, or source address D1_2 and destination address P1_2. In the reverse direction, the same set of IP address pairs should be maintained e.g. the response to a UDP datagram with source address D1_1 and destination address P1_1 will have source address P1_1 (not P1_2) and destination address D1_1. If Network 1 fails then Network 2 can carry all queries and responses (the DSC must set the destination IP of queries to P1_2 once the failure of P1_1 has been detected).

The PSX Connections are implemented using connected UDP sockets, so a UDP packet is read by the NGPW only if the source and destination addresses correspond to those used in the policy request sent to the PSX.

PSX Multihoming Configuration