122 5 SDN and NFV in 5G
5.2.1.5 Summary of Dataplane Functions
Figure 5.8 depicts the details of the forwarding steps in the data path.
When a packet arrives, using its input port, a lookup in the direction
table determines if this packet needs to be classified at this hop or
should be just transmitted to the next hop. If it is the latter, the packet
is sent out according to its MAC address by performing a lookup in
the MAC table. If a classification is needed at this hop to determine its
service sets, then lookups are performed at the Microflow, subscriber,
and application tables. Each table can independently operate on the
service set. The lookup keys are specified in the brackets below the
table names in the figure. According to the bits set in the service set,
the path status table determines which services have already been traversed, and what is the next service in the chain. Finally, the packet is
set with the right MAC address in the next destination table and sent
out to the corresponding output port.
5.2.2 SFC Monitoring
Regardless of what mechanism is used to implement the service chaining, one important problem is how to verify that the path has been
correctly installed. The goal is to prove that packets of a given flow
have traversed the expected path. Existing reachability measurement
includes ping and traceroute to measure the reachability from a source
to a destination. Ping triggers ICMP replies and traceroute triggers
ICMP TTL expiration messages on the routers along the path. Both
methods do not require two-end control. There has been ping and
traceroute at different protocol layers, for example, MPLS ping.
However, as stated earlier, the traditional ping/traceroute method is
not suitable for the inline service setting. In traditional network, the
loss of ping/traceroute packets indicates the path problem. However,
in our setting, the ping/traceroute packet may not be recognized by the
service (middlebox) in the middle of the path, and thus got dropped.
Thus, we cannot simply say the symptom of lost measurement packets
is due to the path reachability problem. Therefore, we need a different
method to measure the path reachability for inline service chaining.
We define the reachability problem of inline service chaining as follows. Assuming that a flow f traverses service chain (S1, S2, and S3) in
order, the network topology is shown in Figure 5.9. In this example,
the services are connected to switches F1, F2, and F3, respectively.
The flow enters from the ingress switch F0 and exits the network from
5.2.1.5 Summary of Dataplane Functions
Figure 5.8 depicts the details of the forwarding steps in the data path.
When a packet arrives, using its input port, a lookup in the direction
table determines if this packet needs to be classified at this hop or
should be just transmitted to the next hop. If it is the latter, the packet
is sent out according to its MAC address by performing a lookup in
the MAC table. If a classification is needed at this hop to determine its
service sets, then lookups are performed at the Microflow, subscriber,
and application tables. Each table can independently operate on the
service set. The lookup keys are specified in the brackets below the
table names in the figure. According to the bits set in the service set,
the path status table determines which services have already been traversed, and what is the next service in the chain. Finally, the packet is
set with the right MAC address in the next destination table and sent
out to the corresponding output port.
5.2.2 SFC Monitoring
Regardless of what mechanism is used to implement the service chaining, one important problem is how to verify that the path has been
correctly installed. The goal is to prove that packets of a given flow
have traversed the expected path. Existing reachability measurement
includes ping and traceroute to measure the reachability from a source
to a destination. Ping triggers ICMP replies and traceroute triggers
ICMP TTL expiration messages on the routers along the path. Both
methods do not require two-end control. There has been ping and
traceroute at different protocol layers, for example, MPLS ping.
However, as stated earlier, the traditional ping/traceroute method is
not suitable for the inline service setting. In traditional network, the
loss of ping/traceroute packets indicates the path problem. However,
in our setting, the ping/traceroute packet may not be recognized by the
service (middlebox) in the middle of the path, and thus got dropped.
Thus, we cannot simply say the symptom of lost measurement packets
is due to the path reachability problem. Therefore, we need a different
method to measure the path reachability for inline service chaining.
We define the reachability problem of inline service chaining as follows. Assuming that a flow f traverses service chain (S1, S2, and S3) in
order, the network topology is shown in Figure 5.9. In this example,
the services are connected to switches F1, F2, and F3, respectively.
The flow enters from the ingress switch F0 and exits the network from