MobilityFirst NP Project Update

NSF PI Meeting

Washington D.C.

May 19-20, 2014

D. Raychaudhuri

WINLAB, Rutgers University

ray@winlab.rutgers.edu

Introduction

WINLAB

MobilityFirst Project: Background

MobilityFirst project started in 2010 under NSF FIA,

continuing under FIA-NP

Project team: Rutgers, UMass, Michigan, Wisconsin,

Duke, MIT, Nebraska

Clean-slate architecture motivated by fundamental shift

of Internet services to mobile platforms ~10B in 2020!

Use cases:

Mobile Data

(cellular, hetnet)

Vehicular Networks Content Delivery

Cloud Services

Internet-of-Things Emergency Networks

WINLAB

MobilityFirst Concepts: Architecture Summary

Routers with Integrated

Storage & Computing Heterogeneous

Wireless Access

End-Point mobility

with multi-homing In-network

content cache

Network Mobility &

Disconnected Mode

Hop-by-hop

file transport Edge-aware

Inter-domain

routing

Named devices, content,

and context

11001101011100100…0011

Public Key Based

Global Identifier (GUID)

Storage-aware

Intra-domain

routing

Service API with

unicast, multi-homing,

mcast, anycast, content

query, etc.

Strong authentication, privacy

Ad-hoc p2p

mode

Human-readable

name

Connectionless Packet Switched Network

with hybrid name/address routing

MobilityFirst Protocol Design Goals: - 10B+ mobile/wireless devices

- Mobility as a basic service

- BW variation & disconnection tolerance

- Ad-hoc edge networks & network mobility

- Multihoming, multipath, multicast

- Content & context-aware services

- Strong security/trust and privacy model

Global Name

Resolution Service

WINLAB

MobilityFirst Concepts: Protocol Stack

IP

Hop-by-Hop Block Transfer

Link Layer 1

(802.11)

Link Layer 2

(LTE)

Link Layer 3

(Ethernet)

Link Layer 4

(SONET)

Link Layer 5

(etc.)

GSTAR Routing MF Inter-Domain

E2E TP1 E2E TP2 E2E TP3 E2E TP4

App 1 App 2 App 3 App 4

GUID Service Layer Narrow Waist GNRS

MF Routing

Control Protocol

NCS Name

Certification

& Assignment

Service

Global Name

Resolution

Service

Data Plane Control Plane

Socket API

Switching

Option

Optional Compute

Layer

Plug-In A

WINLAB

MobilityFirst Concepts: Name-Address

Separation GUIDs Separation of names (ID) from

network addresses (NA)

Globally unique name (GUID)

for network attached objects User name, device ID, content, context,

AS name, and so on

Multiple domain-specific naming

services

Global Name Resolution Service

for GUID NA mappings

Hybrid GUID/NA approach Both name/address headers in PDU

“Fast path” when NA is available

GUID resolution, late binding option

Globally Unique Flat Identifier (GUID)

John’s _laptop_1

Sue’s_mobile_2

Server_1234

Sensor@XYZ

Media File_ABC

Host

Naming

Service

Network

Sensor

Naming

Service

Content

Naming

Service

Global Name Resolution Service

Network address

Net1.local_ID

Net2.local_ID

Context

Naming

Service

Taxis in NB

WINLAB

MobilityFirst Concepts: GUID Service

Example

MobilityFirst Network

(Data Plane)

GNRS

Register “John Smith22’s devices” with NCS

GUID lookup

from directory

GUID assigned

GUID = 11011..011

Represents network

object with 2 devices

Send (GUID = 11011..011, SID=01, data)

Send (GUID = 11011..011, SID=01, NA99, NA32, data)

GUID <-> NA lookup

NA99

NA32

GNRS update

(after link-layer association)

DATA

SID

NAs

Packet sent out by host

GNRS query

GUID

Service API capabilities:

- send (GUID, options, data)

Options = anycast, mcast, time, ..

- get (content_GUID, options)

Options = nearest, all, ..

Name Certification

Services (NCS)

WINLAB

MobilityFirst Concepts: An Example

Context-Aware GeoTag Application

Context: Location L; messages dropped at L; phones that dropped message at L

Location L is a geo fence, and captured by a context GUID: GL

GNRS mappings for GL enable advanced messaging operations Example 1 : ‘Send message to Location L’ :

received by all phones bound to location L

Example 2 : ‘Get messages dropped for Location L’ : requested received by all phones that dropped a message at L

Not an overlay or hosted service

Devices: HTC Evo 4G, Samsung 4G (GSII/Epic Touch), with WiFi and WiMAX radios

Software: Android 2.3.x/4.x, MobilityFirst Protocol Stack and Network Service API

Drop Message

@L

Pickup Messages

@L

WINLAB

MobilityFirst Project: Status Summary

Architecture, protocol spec and design of key

components completed during first phase of FIA Hybrid name/address based network

Public key GUIDs for named objects

Multiple name certification services (no single root of trust)

Global name resolution service (GNRS)

Storage-aware & edge-aware routing protocols

Routers with storage and late binding

Hop-by-hop transport

Simulation and experimental evaluation of major design

elements in MF architecture (GNRS, routing, etc.)

Sample applications developed Content retrieval, context-aware messaging, geo-tagged multicast, ..

WINLAB

MobilityFirst Project: Status Summary (cont.)

Focus on selected use cases: Mobile data services via cellular and hetnet (WiFi + cellular)

Vehicular networks (V2V, V2I)

Content delivery networks (CDN)

Internet-of-things (IoT)

Extensive prototyping and technology evaluation Validation of key components: GNRS, routing, transport, …

Router implementation on various platforms: Click, OpenFlow, FPGA

Two GNRS options prototyped: Dmap & Auspice (deployed on Amazon)

Development of socket API software for Android, Linux platforms

MF cod releases v1.0, 2.0, …

Wide-area experimental deployment on GENI network

Demos of example advanced services such as multi-homing, context-aware

service and named content retrieval

Research collaborations and initial industry engagement

WINLAB

MobilityFirst Project: NP Scope

Overall project focus on: design improvements to respond to recent technology changes and

emerging service needs

Real-world trials for further validation and early adoption

Research themes for NP project: 2.1 Mobile cloud services & compute layer

2.2 Context aware services

2.3 Cellular-Internet convergence

2.4 Technology platforms

2.5 Advanced GNS Services

2.6 Content services

2.7 Security & Privacy

WINLAB

MobilityFirst Project: NP Scope (cont.)

Network Environment (NE) Trials: Continued deployment of long-running MF service slice on GENI

NE1: WHYY/PBS content networking trial over PennREN network

NE2: 5Nines mobile services and cloud trial in Madison, WI

NE3: CASA context-aware emergency service trial in TX

Real-world use, early adoption & standards Evolutionary strategies for MF deployment over existing IP & cellular

networks

Application to key verticals – mobile data, IoT, emergency response

Pre-standards contributions: ICN, IoT, 5G, ..

NP Research Themes

WINLAB

NP Research Themes: Cellular-Internet

Convergence – Current Mobility Solution

High access network cost, gateway bottlenecks,

latency, protocol interworking complexity…

+

-

Fine-grained control over quality of service

WINLAB

NP Research Themes: Cellular-Internet

Convergence – Future Mobility-Centric Internet

Interchangeable access tech, low latency,

improved scalability, advanced mobile services

… Needs a new global networking standard in long run

+

-

WINLAB

• Parametric analysis of best interface vs. dual homing

• Link delay, data rate and download size varied

• Necessity of a soft threshold to stripe across both interfaces or use the best

NP Research Themes: Dual-Homed

Cellular Service via MF

-122.43 -122.42 -122.41 -122.4 -122.39 -122.38 -122.37

37.77

37.775

37.78

37.785

37.79

37.795

37.8

Longitude

La

titid

e

Free Wi-Fi hotspots(AT&T HotSpot Locator)

Simulation of San-Francisco cabs for Wi-Fi /LTE dual-homing

1 2 3 4 50

10

20

30

40

50

60

70

Ave

rag

e th

rou

gh

pu

t p

er

se

c (

in M

bp

s)

Cab no.1 2 3 4 5

0

10

20

30

40

50

60

70

Cab no.

Ma

xim

um

th

rou

gh

pu

t p

er

se

c (

in M

bp

s)

Using only LTE

Using the best available Wi-Fi

Using all the available WiFis

Using all the Wi-Fis and LTE

Only Wi-Fidoes not helpon an average

WINLAB

NP Research Themes: Content Service in

MF – GNRS assisted caching

Popularity based caching using GNRS

GNRS: maintains all the locations of a content

LNRS: maintains local content location and access count of queried

content. manages popularity cache

17

Global Name Resolution Service (GNRS)

LNRS

LNRS

AS1 AS2

host1

abc

cache tree

GUID Local

address

Access

count flag

abc cache

proxy 300 cache

abc

WINLAB

NP Research Themes: Content Service in

MF – Optional Compute Layer Processing

Content cache at mobile

Operator’s network – NA99

User mobility

Content Owner’s

Server

GUID=13247..99

GUID=13247..99 GUID=13247..99

GUID=13247..99

GNRS query

Returns list:

NA99,31,22,43

NA22

NA31

NA99

NA29

NA43

Data fetch from

NA99

Data fetch from

NA43

GNRS

Query

Get (content_GUID,

SID=128 - cache service)

Get (content_GUID)

Enhanced service example – content delivery with in-network caching & transcoding

MF Compute Layer

with Content Cache

Service plug-in

Query

SID=128 (enhanced service) GUID=13247..99

Filter on

SID=128

Mobile’s GUID

Content file

WINLAB

A proof-of-concept prototype

• An extensible Click element for the MobilityFirst router

• Services: protocol translator, WAN optimizer, firewall, confidential communication modules…

Micro-benchmarking

• Evaluation on Deterlab/Emulab

• Small additional delay for a computationally intensive service

• Scale well to handle high-throughput data traffic

Real-world Deployment

• Evaluation on PlanetLab (nodes connected via virtual point-to-point links)

• A series of Internet-based evaluations are on-going

NP Research Themes: PacketCloud

Implementation for MF Compute Layer

19

WINLAB

NP Research Themes: Mobile Cloud

Goal: bring cloud resources to mobiles transparently 1. Improve mobile computation speed & latency

2. Require no programmer effort

3. Generalize well with existing applications

4. Resist network failures

20

Variable latency,

bandwidth

Leverage

multicore servers

High-speed, low-latency network

to application servers

Multiple groups involved at:

Umass, Michigan, Duke, Rutgers

WINLAB

NP Research Themes: Mobile Cloud (cont.)

If future cloud services will largely serve mobile devices

What new functions are necessary in cloud layer for

hosted applications to better service mobile clients?

Techniques for mobile cloud Edge: Compute resources at edge network to minimize

latency

Geo* : geo-location, geo-replication, geo-elastic, …

Dynamic migration: move cloud service state to follow

user mobility

2

1

WINLAB

NP Research Themes: Mobile Cloud Using MF

Service Addressability and Anycast

Technical Approach: MF service Addressability via GUID, anycast, virtual networks, ..

WINLAB

NP Research Themes: Technology

Platforms – Click Software Router

Inter-Domain (EIR)

Multicast

Lightweight, scalable multicast • GNRS for maintenance of

multicast memberships

• Heuristic approaches to

reduce network load, limit

duplicated buffering, and

improve aggregate delivery

delays

• Click prototype, with SID for

multicast flows

• Evaluating hail a cab

application as a example

multipoint delivery scenario

In the logically central view, the MF controller manages local discovery of hosts, storage and compute resources, and implements storage aware routing (GSTAR) to handle host mobility

To handle inter-network mobility, controller interfaces with GNRS service plane to forward packets on inter-domain paths

OpenFlow-based prototype with Floodlight controller Implementation with OpenDaylight

controller and OF v1.3 in progress

GUID being mapped to IPv6 addr for direct matching within switch

Performance comparison of forwarding ops: GUID, NA , late-binding (GNRS) Penalty is severe for controller

interaction required for new destination transfers and to handle mobility events

Proactive approaches to limit switch-controller communication being explored

NP Research Themes: Technology Platforms

- OpenFlow

5/28/2014 WINLAB, Rutgers University

24

virtual

router

In-network

storage for

store/forwar

d

Local compute

cloud

Global Name Services

MF SDN

Controller

4G/WiMAX BTS

WINLAB

NP Research Themes: Technology

Platforms – FLARE SDN (U Tokyo)

Objectives Multi-site deployment of MobilityFirst

routing and name resolution services

Impact of large RTTs on MobilityFirst

network protocols

High performance evaluation of

MobilityFirst delivery services on

FLARE - 1Gbps, 10Gbps

Augmented Click router elements

compiled down to FLARE native

Evaluation of FLARE platform for

design and evaluation of next-

generation network protocols

Demo at GEC-16, March 2013

NP Research Themes: Technology

Platforms - GENI Deployed MF components: software

router (Click-based), name resolvers, clients with MF host stacks

GENI resources: multi-site layer-2 network on Internet 2 and NLR, programmable raw and VM nodes at ProtoGENI and GENI rack sites, WiMAX campus deployments

Core network operation, visualization, and analysis from Rutgers, WINLAB

Applications/clients can connect to MF network from any GENI WiMAX access sites

WiFi and Ethernet connectivity for clients can be enabled with little effort

5/28/2014 WINLAB, Rutgers University

26

NP Network

Environments

WINLAB

NE Trials:

MobilityFirst being validated experimentally via wide-

area GENI trials with 10’s of nodes nationwide and ~1-2

access networks (WiMax, WiFi)

Next step technology and service trials with early

adopters in real-world settings PBS/WHYY content service trial in PA

5Nines mobile data and cloud service trial in Madison, WI

CASA context-aware emergency messaging trial in TX

WINLAB

NE Trials: MF Multi-Site Deployment

on GENI

Salt Lake, UT

Cambridge,

MA

N. Brunswick,

NJ

Ann Arbor, MI Madison, WI

Tokyo, Japan

Lincoln, NE

Los Angeles,

CA Clemson,

SC

Long-term (non-

GENI)

MobilityFirst Access

Net

Short-term

Wide Area ProtoGENI

Palo Alto, CA

ProtoGENI

MobilityFirst

Routing and Name

Resolution

Service Sites

I2

NL

R

Atlanta, GA

WINLAB

NE Trials: GENI Deployment of MobilityFirst

at GEC18, Oct 2013 MF Routing and Naming

Services deployed at 5 GENI rack sites with Internet2’s AL2S providing cross-site layer-2 connectivity

Rutgers and NYU Poly (with rack at NYU) routers connected to WiFi and WiMAX access networks

Android phones with WiFi/WiMAX connectivity ran MF stack and demo application (Drop It)

5/28/2014 WINLAB, Rutgers University 30

Wisconsin

GENI rack

Utah

GEN

I

rack

BBN

GENI

rack GENI Internet2

Core

GENI

Edge

GENI

Edge

WiMAX

BTS

WiMAX

BTS

MobilityFirst

Software Router

with GNRS

Dual homed

Android phone

with WiFi/WiMAX

with MF stack

ORBIT radio

node with WiFi

as MF AP

Sample “GeoTag”

context-aware

application

WINLAB

NE Trials: 5Nines Network, Madison

NE1: Mobile Data Service Trial with 5Nines (ISP) in Madison, WI

WINLAB

5nines access network

Other sites

Internet

WiMAX

Base station

Compute

resource

WiFi

AP

Compute

resource

UW enterprise network

5nines data

center and NAP

User devices

Vehicular

clients

Logical structure of the various WiMAX and WiFi infrastructure to be used in experiments

Central ASN and Radius service

NE Trials: 5Nines Network Topology

WINLAB 33

User1

User2

Internet

Media transcoding

as an edge computing

service

Original source

Sample “mobile cloud” service scenario:

Wireless bit-rate adaptive media transcoding in the network

HD

Mid Mid

Low Res

NE Trials: Cloud Transcoding App in

5Nines Network

WINLAB

NE Trials: PBS/WHYY Content

Service Trial in PA

NE2: Content Services Trial with PennREN/PBS in PA

WINLAB

NE Trials: PBS Services Summary

Multiple regional broadcast stations with video and radio programming Programs from national (PBS), regional and local sources interleaved

High bandwidth connectivity between stations PennREN - New fiber network

Satellite feed for receiving national programs

At each station, multiple content edit and distribution workflows …. … before individual programs are sequenced for on-air transmission

Content is also pushed for Web distribution (e.g., MOV format)

Multiple channels of distribution (e.g., CDNs, PBS website, WHYY website)

Challenges in content management, retrieval, workflow orchestration, and distribution to different channels with differing requirements

5/28/2014 WINLAB, Rutgers University 35

WINLAB

Content Naming and Resolver Services Stations are looking to ‘package’ their content (both archived and current) and make it

available to interested parties over the Web

MF can support scalable resolution of content GUIDs and metadata searches through Global Naming Service (GNS), even under content mobility – can enable store anywhere solutions

Multipoint delivery protocols Content is moved (inefficiently) to multiple channels for storage, sequencing, and distribution

MF transport protocols with multipoint delivery (multicast, multipath/multihome) can improve efficiency

MF segmented data transport can improve Web delivery to wireless and multihomed devices

Store/Compute and Forward Workflows (often) involve automated compute segments before processed content is

delivered. E.g., Content is placed in a ‘watch‘ folder while closed captions are obtained from external processes

MF compute layer extensibility can enable in-network integration of the compute components for these workflows during end-to-end transport

MF can similarly support in-network transcoding options, and caching solutions for end-user delivery

5/28/2014 WINLAB, Rutgers University 36

NE Trials: MF Features in PBS Trial

WINLAB

NE Trials: CASA Emergency Network

NE3: Context-Aware Emergency Notification System (CASA)

WINLAB

Resources

Project website: http://mobilityfirst.winlab.rutgers.edu

GENI website: www.geni.net

ORBIT website: www.orbit-lab.org

WINLAB website: www.winlab.rutgers.edu