fog$networks$ - 5g summit · whatis$“fog$network”?$$ • a$network$architecture$that...

Post on 22-Jul-2020

1 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Fog  Networks  

 Mung  Chiang  

Princeton  University  2015  

From  Cloud  to  Fog  

2000  –  2015     2015  –  2030  ?      

What  is  “Fog  Network”?    

•  A  network  architecture  that  uses  one  or  a  collaboraGve  mulGtude  of  end-­‐user  clients  or  near-­‐user  edge  devices  to  carry  out  a  substanGal  amount  of  storage  (rather  than  stored  primarily  in  cloud  data  centers),  communicaGon  (rather  than  routed  over  backbone  networks),  and  control,  configuraGon,  measurement  and  management  (rather  than  controlled  primarily  by  network  gateways  such  as  those  in  LTE  core).    

Many  Types  of  Clients  &  Edge  Devices  

Contrast  Them  With…  

Client/Edge/IoT  Devices  

•  Are  recently  becoming    –  Powerful  (in  sensing,  storage,  compuGng,  control,  comm.)  –  SGll  limited  (in  baYery,  storage,  compuGng,  informaGon)  – Maybe  mobile  

•  Crowds  of  clients/edge  devices  are    –  Dense    –  Distributed  –  Under-­‐organized  

It  has  become  both  feasible  and  interesGng  to  ask:    

“Can  ‘this’  be  done  at  clients/edge?”          

Rather  Than:  

1.  LTE  Core  (P-­‐GW,  S-­‐GW,  PCRF…)  2.  Data  Center    3.  Wide  area  backbone    

TradiGonal  View  

use    

Fog  View  

are  (part  of)  

What  If…  •  The  set-­‐top  box  in  your  living  room  replaces  the  DPI  box?    •  The  dashboard  in  your  car  is  your  cloud  caching  content?  •  Your  phone  (and  other  phones)  become  LTE  PDN-­‐GW  &  PCRF?    

•  The  “network  edge”  gives  you  the  edge  •  The  clients  are  the  controllers  

Why  Fog?  

Brick  vs.  Click:  Physical  interacGon  &  real  estate    

1.  Real-­‐Gme  processing:  right  here  and  now  2.  Rapid  innovaGon  and  affordable  scaling      3.  Client-­‐centric  objecGves/privacy    4.  Pooling  of  local  content/BW/storage/processor  5.  Operate  on  encrypted  and  mulGpath  traffic  

Case:  OTT  Smart  Data  Pricing  (SDP)  

Clients  can  crowd-­‐source  network  inference/measurement  and  overlay  billing  

PCRF%

SGSN%

GGSN%

PCEF%

MSP%

OTT%Pricing%System%

Client%%Device%

ISP%Connector%Client%Connector%

Pricing%Policy%&%Algorithms%

SDK  

OTT  

Case:  Client-­‐controlled  Cloud  Storage  

Google Drive!

Dropbox!Enterprise

server!

CYRUS! CYRUS!

Box!

Client Control!

CYRUS!

Client Control!

Google Drive!

Client/edge  intelligence  can  commodiGze  cloud  resources  

Case:  Client-­‐Side  HetNets  Control  

Unlicensed   Licensed,  Planned   Licensed,  Unplanned  

Core  Network  

Control  Plane

 Da

ta  Plane

 

Internet  

RNS  (RNC,  eNodeB)  

Wi-­‐Fi  AP  

HNS  (SeGW,  HNB-­‐GW,  HomeNodeB)  

Clients  can  autonomously  manage/control  their  own  configuraGons  

Case:  Client  Bandwidth  Pooling  

Idle  resources  on  the  edge  pooled  but  shared  unpredictably      

Fog  as  an  Architecture  

EDD  Edge-­‐Driven  “Data-­‐center”  

EDC  Edge-­‐Driven  Control-­‐plane  

5G  Mobile     IoT     Data  analyGcs    

Examples  •  Over  The  Top  (OTT)  content  management  and  SDP  •  Client-­‐centric  HetNets  Control    •  Client-­‐controlled  cloud  storage  •  Session  management  and  signaling  load  opGmizaGon  •  Crowd-­‐sensing  inference  of  network  states  •  Edge  analyGcs  and  real-­‐Gme  stream-­‐mining    

•  Clients’  idle  compuGng/storage/content  pooling  •  Edge  caching/BW  management  at  home  gateway/small  cell    •  Client-­‐driven  distributed  beam-­‐forming    •  FlashLinQ/LTE  Direct/WiFi  Direct/AirDrop  •  Cloudlets/Micro  Datacenters/Mobile  CDN  

Cloud  –  Fog  Interface  

Massive  storage  Heavy  duty  computaGon  Global  coordinaGon  Wide-­‐area  connecGvity    

Real  Gme  processing  Rapid  innovaGon  Client-­‐centric  Edge  resource  pooling    

Networking  Revisited  

Relocate  Fog    

Virtualize  SDN  

Redefine  CCN  

Research  Challenges  •  Trustworthiness  /  verificaGon  of  client/edge  solware  &  hardware  

•  IncenGvizaGon  of  client  parGcipaGon  

•  InteracGons  w.  client/IoT  OS  and  definiGon  of  network  service  APIs    •  Interfaces:  Cloud-­‐to-­‐cloud  and  cloud-­‐to-­‐fog    •  OscillaGon/divergence  and  global  configuraGon  consistency  during  

the  interacGons  of  local  acGons  

•  Tradeoff  of  Local  vs.  Global  architecture,  Distributed  vs.  Centralized    

Inter-­‐Disciplinary  SoluGons  

Network  Engineering    

Device  Hardware/OS  

Economics  &  Pricing  

HCI  &    App  UI/UX  

Data  Science  

hYp://fogresearch.org  

 “Should  ‘this’  be  done  at  clients/edge?”    

     

Impact  on  Value  ProposiGon  along  Ecosystem  Food-­‐chain  

•  End  user  experience  providers?    •  Network  operators?    •  Equipment  vendors?    •  Cloud  service  providers?    •  System  integrators?    •  Edge  device  manufacturers?    •  Client/IoT  device  manufacturers/OS?    •  Chip  suppliers?    

     

Industry-­‐Academia  CollaboraGon          

 chiangm@princeton.edu  

 

top related