emulation

Emulation environment

Start a topology

1. sudo python place_emu/emulator/topology_zoo.py -g inputs/networks/Abilene.graphml
2. Check if everything is working (other terminal): vim-emu datacenter list

Start and place a service (example)

Used service

The used service has four containers: User -> Proxy (Squid) -> L4FW (Socat) -> Webservice (Apache)

     Example Service Chain an IPs

       +-------+        +------+
       |       |        |      |
       |  User |        | Web  |
       |       |        |      |
       +-------+        +---^--+
10.0.0.1/24|                | 30.0.0.2/24
           |                |
10.0.0.2/24|                | 30.0.0.1/24
       +---v---+        +---+--+
       |       |        |      |
       | Proxy +--------> L4FW |
       |       |        |      |
       +-------+        +------+
         20.0.0.1/24  20.0.0.2/24

Deployment steps

# deploy VNFs
vim-emu compute start -n vnf_user -i placement-user-img --net '(id=output,ip=10.0.0.1/24)' -d pop0
vim-emu compute start -n vnf_proxy -i placement-squid-img --net '(id=input,ip=10.0.0.2/24),(id=output,ip=20.0.0.1/24)' -d pop1
vim-emu compute start -n vnf_l4fw -i placement-socat-img --net '(id=input,ip=20.0.0.2/24),(id=output,ip=30.0.0.1/24)' -d pop2
vim-emu compute start -n vnf_web -i placement-apache-img --net '(id=input,ip=30.0.0.2/24)' -d pop3

# setup network forwarding rules
vim-emu network add -src vnf_user:output -dst vnf_proxy:input
vim-emu network add -src vnf_proxy:output -dst vnf_l4fw:input
vim-emu network add -src vnf_l4fw:output -dst vnf_web:input

(this script is also available in emulator/deploy_example.sh)

Testing the deployment

# check deployment
vim-emu compute list
+--------------+-------------+----------------------+------------------+-------------------------+
| Datacenter   | Container   | Image                | Interface list   | Datacenter interfaces   |
+==============+=============+======================+==================+=========================+
| pop3         | vnf_web     | placement-apache-img | input            | dc4.s1-eth3             |
+--------------+-------------+----------------------+------------------+-------------------------+
| pop2         | vnf_l4fw    | placement-socat-img  | input,output     | dc3.s1-eth3,dc3.s1-eth4 |
+--------------+-------------+----------------------+------------------+-------------------------+
| pop1         | vnf_proxy   | placement-squid-img  | input,output     | dc2.s1-eth3,dc2.s1-eth4 |
+--------------+-------------+----------------------+------------------+-------------------------+
| pop0         | vnf_user    | placement-user-img   | output           | dc1.s1-eth3             |
+--------------+-------------+----------------------+------------------+-------------------------+

# check basic connectivity
containernet> vnf_user ping -c3 10.0.0.2
containernet> vnf_proxy ping -c3 20.0.0.2
containernet> vnf_l4fw ping -c3 30.0.0.2

# full chain HTTP connectivity
containernet> vnf_user curl -x http://10.0.0.2:3128 http://20.0.0.2:8899

This curl command might look confusing. It does the following:

• a HTTP request to the IP of vnf_l4fw that forwards the request arriving at port TCP:8899 to vnf_web on port TCP:80
• for the request it has to use the proxy (vnf_proxy) which is specified by -x http://10.0.0.2:3128

Manual RTT measurement

(in a new terminal)

# log into to vnf_user
docker exec -it mn.vnf_user /bin/bash

# basic check
vnf_user curl -x http://10.0.0.2:3128 http://20.0.0.2:8899 -v

# use httping to measure RTT between user and vnf_web
vnf_user httping --proxy 10.0.0.2:3128 --url http://20.0.0.2 -p 8899 -c 5

Shut down experiment

containernet> exit

Notes

• Each node of the given graph is turned into an emulated datacenter (PoP)
• PoPs are named based on the node IDs of the networkx graph (not their labels): pop<ID> e.g. pop42
• Link latencies are calculated based on node's geo. locations
• Link bandwidth is set if given in the graph (not many topologies have it)