Intelligent resource placement module for 5GCity distributed cloud and radio platform.
The main goal of the algorith is to take into consideration the network topology and the access points locations in determining the best option for placing the VNFs that compose a Network Service.
- Network Service (NS) placement request
- Device (Network) graph - a description of the available nodes (data-center, edge) and the interconnections between them
- Available resources per device - CPU, Memory, Storage
- Connection links capacity
- NS user gateway - the end node where the NS user is connected
In 5GCity, we can assume that end users of a NS are the SCs and the WiFi APs and the NS gateway is the edge or DC node to which they have a direct connection
The output of the module is one or a group of devices where the NS should be placed.
POST Endpoint http://localhost:3800/placement/
Triggers the placement algorithm. The needed input information is extracted from the slice data (compute and network chunks, resources, links).
- Asynchronous, returns an id of the request.
- The result can be obtained by GET /placement/{placement-id}
{
"fwchain": {
"source": [
{
"node": "pop0",
"vnf": "vnf_user",
"flows": [
{
"data_rate": 1,
"id": "f1"
}
]
}
],
"vnfs": [
{
"name": "vnf_user",
"inputs_fwd": 0,
"inputs_bwd": 0,
"outputs_fwd": 1,
"outputs_bwd": 0,
"out_fwd": [],
"out_bwd": [],
"mem": [
2
],
"cpu": [
2
],
"vnf_delay": 0,
"stateful": false,
"type": "source"
},
{
"name": "vnf_fw1",
"inputs_fwd": 1,
"inputs_bwd": 0,
"outputs_fwd": 1,
"outputs_bwd": 0,
"out_fwd": [
[
1,
0
]
],
"out_bwd": [],
"mem": [
4,
32
],
"cpu": [
8,
16
],
"vnf_delay": 35,
"stateful": false,
"type": "normal"
},
{
"name": "vnf_web",
"inputs_fwd": 1,
"inputs_bwd": 0,
"outputs_fwd": 0,
"outputs_bwd": 0,
"mem": [
8,
0
],
"cpu": [
8,
0
],
"out_fwd": [],
"out_bwd": [],
"vnf_delay": 20,
"stateful": false,
"type": "normal"
}
],
"name": "fw1chain",
"vlinks": [
{
"dest": "vnf_fw1",
"src": "vnf_user",
"direction": "forward",
"dest_input": 0,
"src_output": 0,
"max_delay": 50
},
{
"dest": "vnf_web",
"src": "vnf_fw1",
"direction": "forward",
"dest_input": 0,
"src_output": 0,
"max_delay": 50
}
]
},
"network": {
"directed": false,
"graph": {
"node_default": {},
"edge_default": {}
},
"nodes": [
{
"mem": 2,
"cpu": 2,
"id": 0
},
{
"mem": 4,
"cpu": 16,
"id": 1
},
{
"mem": 8,
"cpu": 32,
"id": 2
},
{
"mem": 8,
"cpu": 32,
"id": 3
}
],
"links": [
{
"source": 0,
"target": 1
},
{
"source": 0,
"target": 2
},
{
"source": 1,
"target": 3
}
],
"multigraph": false
}
}
- {placement-id}: id of the placement request as body
- Response codes: 202 Accepted, 400 Bad Request
GET Endpoint http://localhost:3800/placement/{placement-id}
Body example
{
"placement": {
"vnfs": [
{
"name": "vnf_fw1",
"node": "pop1"
},
{
"name": "vnf_user",
"node": "pop0"
},
{
"name": "vnf_web",
"node": "pop3"
}
],
"vlinks": [
{
"src_vnf": "vnf_fw1",
"src_node": "pop1",
"dest_vnf": "vnf_web",
"dest_node": "pop3"
},
{
"src_vnf": "vnf_user",
"src_node": "pop0",
"dest_vnf": "vnf_fw1",
"dest_node": "pop1"
}
],
"cpu_oversub": [],
"mem_oversub": [],
"dr_oversub": [],
"alloc_node_res": [
{
"name": "vnf_fw1",
"node": "pop1",
"cpu": 8,
"mem": 4
},
{
"name": "vnf_user",
"node": "pop0",
"cpu": 2,
"mem": 2
},
{
"name": "vnf_web",
"node": "pop3",
"cpu": 8,
"mem": 8
}
],
"flows": [
{
"arc": "vnf_fw1.0->vnf_web.0",
"src_node": "pop1",
"dst_node": "pop3",
"src_vnf": "vnf_fw1",
"dest_vnf": "vnf_web",
"flow_id": "f1"
},
{
"arc": "vnf_user.0->vnf_fw1.0",
"src_node": "pop0",
"dst_node": "pop1",
"src_vnf": "vnf_user",
"dest_vnf": "vnf_fw1",
"flow_id": "f1"
}
],
"links": [
{
"arc": "vnf_fw1.0->vnf_web.0",
"edge_src": "pop1",
"edge_dst": "pop3",
"link_src": "pop1",
"link_dst": "pop3"
},
{
"arc": "vnf_user.0->vnf_fw1.0",
"edge_src": "pop0",
"edge_dst": "pop1",
"link_src": "pop0",
"link_dst": "pop1"
}
]
},
"metrics": {
"num_instances": 3,
"max_cpu_oversub": 0,
"max_mem_oversub": 0,
"changed": [
{
"name": "vnf_fw1",
"node": "pop1"
},
{
"name": "vnf_user",
"node": "pop0"
},
{
"name": "vnf_web",
"node": "pop3"
}
],
"num_changed": 3,
"path_delays": [
{
"src": "vnf_fw1",
"dest": "vnf_web",
"src_node": "pop1",
"dest_node": "pop3",
"path_delay": 0
},
{
"src": "vnf_user",
"dest": "vnf_fw1",
"src_node": "pop0",
"dest_node": "pop1",
"path_delay": 0
}
],
"vnf_delays": [
{
"vnf": "vnf_fw1",
"vnf_delay": 35
},
{
"vnf": "vnf_user",
"vnf_delay": 0
},
{
"vnf": "vnf_web",
"vnf_delay": 20
}
],
"total_path_delay": 0,
"total_vnf_delay": 55,
"max_endToEnd_delay": 55,
"total_delay": 55,
"max_dr_oversub": 0
},
"id": "7da2d000-b104-4c52-93b5-64863a01a3b3"
}
Response codes: 200 OK, 404 Item Not Found
python setup.py install
Requires Python 3.5+
Run placement to start the server
Sevil Dräxler, Stefan Schneider, Holger Karl: "Scaling and Placing Bidirectional Services with Stateful Virtual and Physical Network Functions". IEEE Conference on Network Softwarization (NetSoft), Montreal, CA (2018)
"B-JointSP is an optimization problem focusing on the joint scaling and placement (called embedding) of NFV network services, consisting of interconnected virtual network functions (VNFs). The exceptional about B-JointSP is its consideration of realistic, bidirectional network services, in which flows return to their sources. It even supports stateful VNFs, that need to be traversed by the same flows in both upstream and downstream direction. Furthermore, B-JointSP allows the reuse of VNFs across different network services and supports physical network functions."