The network is built and configured with GNS3 ide. It is recommended to use the same software for testing it.
The above scenario shows three networks interconnected through a core network of 5 routers. Each network contains either Client nodes (A1, A2, C1) or Server nodes (B1, B2). Clients send data to servers and have the flow characteristics that are shown in the following table:
| Source | Destination | Priority | Bandwidth |
|---|---|---|---|
| A1 | B1 | High | 0.9Mbps |
| A2 | B2 | Low | 1.2Mbps |
| C1 | B1 | High | 0.9Mbps |
-
Each flow conforms to the characteristics of the above able. For high-priority flows, the excess traffic is downgraded to low priority; for low-priority flows, the excess traffic is dropped.
-
Each high-priority flow is granted at worst 60% of the bandwidth for the link it traverses in the core network.
The following table represents the address assignation for the networks:
| Net | Address | Subnet mask |
|---|---|---|
| A | 10.0.1.0 | 255.255.255.0 |
| B | 10.0.2.0 | 255.255.255.0 |
| C | 10.0.3.0 | 255.255.255.0 |
| R1-R3 | 172.16.0.0 | 255.255.255.252 |
| R1-R2 | 172.16.4.0 | 255.255.255.252 |
| R3-R4 | 172.16.8.0 | 255.255.255.252 |
| R4-R5 | 172.16.12.0 | 255.255.255.252 |
| R2-R5 | 172.16.16.0 | 255.255.255.252 |
For the IPs assigned to the router interfaces you can refer to the picture shown above.
In order to achive the flow characteristics, we decided to configure the IP address of hosts in Network-A and Network-B statically. Instead we configured DHCP service on R2 since it has only one host in its subnetwork.
Futhermore, as shown in the next table, for each router we configured the Loopback0 interface with IP mask of 32 bits, which will be used as router-id attribute in the OSPF protocol.
| Router | Loopback IP address |
|---|---|
| R1 | 172.16.1.1 |
| R2 | 172.16.1.2 |
| R3 | 172.16.1.3 |
| R4 | 172.16.1.4 |
| R5 | 172.16.1.5 |
We chose OSPF as routing protocol since it is ready for the implementation of MPLS-TE.
Every router has been configured in order to be able to announce all the network to which it is attached, included the Loopback0:
R(config)# router ospf 1
R(config-router)# network <network-address> <wildcard-mask> area 0
In order to classificate the flows based on source and destination we used Extendend Access Control Lists and we put them in the nearest router to the source. Thus, we defined three ACLs accordingly to the three different flows we have to match.
On R1 we defined an ACL with id 121 for traffic from A1 to B1:
R1(config)# access-list 121 permit ip host 10.0.1.2 host 10.0.2.2
R1(config)# class-map match-all A1B1
R1(config-cmap)# match access-group 121
and an ACL with id 122 for traffic from A2 to B2:
R1(config)# access-list 122 permit ip host 10.0.1.3 host 10.0.2.3
R1(config)# class-map match-all A2B2
R1(config-cmap)# match access-group 122
On R2 we defined an ACL with id 121 for traffic from C1 to B1:
R2(config)# access-list 121 permit ip host 10.0.3.2 host 10.0.2.2
R2(config)# class-map match-all C1B1
R2(config-cmap)# match access-group 121
For differentiating the two types of flow we assigned different Per Hob Behavior as shown in the following table:
| Priority | PHB | DSCP |
|---|---|---|
| High | EF | 46 |
| Low | AF13 | 14 |
In the relative routers we defined marking for the classes. On R1 for high-priority flow:
R1(config)# policy-map E00
R1(config-pmap)# class A1B1
R1(config-pmap-c)#set ip dscp ef
and for low-priority flow:
R1(config)# policy-map E00
R1(config-pmap)# class A2B2
R1(config-pmap-c)#set ip dscp af13
On R2:
R2(config)# policy-map E00
R2(config-pmap)# class C1B1
R2(config-pmap-c)#set ip dscp ef
Then we mapped in input the defined policies to the ethernet 0/0 interfaces on both routers:
R(config)# interface ethernet 0/0
R(config-if)# service-policy input E00
In order to meet the requirements about the excess traffic we defined policing rules by means of Committed Access Rate (CAR) at the input of R1 and R2 on ethernet 0/0 interface.
On R1 and R2, for A1 to B1 and C1 to B1 traffic, respectively, defined by ACL 121, we have to downgrade the excess traffic to low-priority:
R(config-if)# rate-limit input access-group 121 900000 5000 5000 conform-action continue exceed-action set-dscp-transmit 14
and only on R1 for A2 to B2 traffic defined by ACL 122 we have to drop packets belonging to exceeded traffic:
R1(config-if)# rate-limit input access-group 122 1200000 5000 5000 conform-action continue exceed-action set-dscp-transmit drop
Finally we configured all routers in the network in order to make them able to classify different traffic classes based on DSCP and to allocate the wanted bandwidth for each of them.
Since we have to reserve at least 60% of bandwidth for the high-priority flows we need only to match EF PHB.
For classifying the flow based on DSCP:
R(config)# class-map match-all HIGH
R(config-cmap)# match dscp ef
For allocating 60% of bandwidth:
R(config)# policy-map OUT
R(config-pmap)# class HIGH
R(config-pmap-c)# bandwidth percent 60
For associating the policy to every serial interface:
R(config-if)# service-policy output OUT
Since now, every traffic from Network-A to Network-B and from Network-C to Network-B is passing through R2-R5 link based on OSPF best path. But by requirements, we have to guarantee at least 60% of bandwidth to high-priority flows. If both of them share R2-R5 link at the same moment, this requirement can not be met.
Furthermore the links involving R1-R3-R4-R5 are never used since they represent the longest path for every traffic among the three stub networks.
Thus, for meeting the bandwidth requirement and avoiding the under-utilization of some network parts, we used MPLS-TE which allows us to support constrained-based routing.
First, we enabled Cisco Express Forwarding (CEF) and tag switching on every interface of every router involved, in our case all of them but R2:
R(config)# ip cef
R(config)# interface <interf>
R(config-if)# mpls ip
Then we made routers able to create traffic engineering tunnels (MPLS Label Switched Paths) to steer traffic through the network allowing links not included in the shortes path to be used:
- Enable tunnels creation:
R(config)# mpls traffic-eng tunnels
- Enable TE extension on OSPF:
R(config)# router ospf 1
R(config-router)# mpls traffic-eng area 1
R(config-router)# mpls traffic-eng router-id Loopback0
- Enable MPLS tunnel creation on the interfaces specifying the reservable bandwidth and the largest reservable flow on the interface:
R(config)# interface <interf>
R(config-if)# mpls traffic-eng tunnels
R(config-if)# ip rsvp bandwidth 512 512
Finally, on the ingress router R1, we created Tunnel0 with 172.16.1.5 as destination and the explicit path:
R1(config)# interface Tunnel1
R1(config-if)# ip unnumbered Loopback0
R1(config-if)# tunnel destination 172.16.1.5
R1(config-if)# tunnel mode mpls traffic-eng
R1(config-if)# tunnel mpls traffic-eng autoroute announce
R1(config-if)# tunnel mpls traffic-eng priority 2 2
R1(config-if)# tunnel mpls traffic-eng path-option 1 explicit name longpath
R1(config-if)# tunnel mpls traffic-eng path-option 2 dynamic
The explict path longpath has been defined as follows:
R1(config)# ip explict-path name longpath enable
R1(cfg-ip-expl-path)# next-address 172.16.0.2
R1(cfg-ip-expl-path)# next-address 172.16.0.10
R1(cfg-ip-expl-path)# next-address 172.16.0.14
send the traffic from A2 to B2 through R2 by setting the routing of the specific destination and the next hop in R1.
R1(config)# ip route 10.0.2.3 255.255.255.255 172.16.0.6
Use the following commands on the end-point consoles to test the resulting traffic load.
On servers:
iperf -u -s -i 1
On clients:
iperf -u -c <server-ip-address> -i 1 -b <data-size> -t 30
The <data-size> can be set to various ranges from small to large to see the run time bandwidth of each flow.
In order to clone this repository you need to have the Git Large File Storage installed on your machine. Also you have to instantiate your working directory with git lfs by:
git lfs install
For pulling for the first time the all content of the repo after the clone, you need to issue:
git lfs pull