MobilityFirst: A Robust and Trustworthy Mobility-Centric ...€¦ · MobilityFirst FIA Team...
Transcript of MobilityFirst: A Robust and Trustworthy Mobility-Centric ...€¦ · MobilityFirst FIA Team...
MobilityFirst: A Robust and Trustworthy Mobility-Centric Architecture for the Future InternetNSF FIA MeetingNov 15-16, 2010
Contact: D. [email protected]
MobilityFirst Project Team
MobilityFirst FIA Team Presentation Nov 15, 2010
MobilityFirst Project: Collaborating Institutions
(LEAD)
+ Also industrial R&D collaborations with AT&T Labs,Bell Labs, Technicolor, Toyota ITC, NEC, Ericsson and others
D. Raychaudhuri, M. Gruteser, W. Trappe, R, Martin, Y. Zhang, I. Seskar
S. Bannerjee W. LehrZ. Morley Mao
B. Ramamurthy
G. ChenX. Yang, R. RoyChowdhury
A. Venkataramani, J. Kurose, D. Towsley
M. Reiter
MobilityFirst Vision
MobilityFirst FIA Team Presentation Nov 15, 2010
INTERNETINTERNET
Vision: Mobility as the key driver for the future Internet
Historic shift from PC’s to mobile computing and embedded devices…
~4 B cell phones vs. ~1B Internet-connected PC’s in 2010Mobile data growing exponentially – Cisco white paper predicts >1exabyte per month (surpassing wired PC traffic) by 2012Sensor deployment just starting, ~5-10B units by 2020
WirelessEdge Network
WirelessEdge Network
INTERNETINTERNET
~1B server/PC’s, ~700M smart phones
~2B servers/PC’s, ~10B notebooks, PDA’s, smart phones, sensors
~2010 ~2020
WirelessEdge Network
WirelessEdge Network
MobilityFirst FIA Team Presentation Nov 15, 2010
INTERNETINTERNET
Vision: Near-term “mobile Internet” usage scenario – cellular convergence
~4-5B new cellular devices in just a few years will drive convergence of technical standards and business models
Currently involves 2 sets of addresses (cellular number & IP), 2 sets of protocols (3GPP and IP), and protocol gateways (GGSN, PDN GW, etc.)Scalability, performance and security problems when bridging 2 networksEven more complicated with multiple radio access systemsLack of a single unified standard inhibits mobile Internet app development across diverse networks and platforms
Cellular –Internet
GW
Cellular –Internet
GW MOBILE INTERNET
MOBILE INTERNET
Cellular system A
Cellular system B
Radio Access Net A
Radio Access B
Radio Access C
MobilityFirst FIA Team Presentation Nov 15, 2010
Vision: Near-term “mobile Internet” usage scenario – Mobile P2P and InfostationsP2P and Infostations (DTN-like) modes for content delivery becoming mainstream
Heterogeneous access; network may be disconnected at timesInvolves both terminal and router mobility; dynamic trust establishmentRequires content caching and opportunistic data delivery
Mobile DTN Router
Roadway Sensors
Mobile DTN User/Router
Ad-HocNetwork
OpportunisticHigh-Speed Link(MB/s)
Mobile P2P User
InfostationsRouter
MOBILE INTERNETMOBILE INTERNET
MobilityFirst FIA Team Presentation Nov 15, 2010
Vision: Future “mobile Internet” usage scenarios – vehicular networks
100’s of million cars will be equipped with radios by ~2015Both V2V (vehicle-to-vehicle) and V2I (vehicle-to-infrastructure) modesInvolves capabilities such as location services, georouting, reliable multicastImportant new trust (security and privacy) requirements in this scenario
Irrelevant vehicles in radio range for few seconds
Passing vehicle,in radio range for seconds
Following vehicle,in radio range for minutes
V2I
V2V
MobilityFirst FIA Team Presentation Nov 15, 2010
Vision: Emerging “mobile Internet” usage scenarios – pervasive (ubiquitous) systems
The next generation of Internet applications will involve interfacing human beings with the physical world
Wide range of usage scenarios including healthcare, smart grids, etc.Networking requires awareness of location and context, along with embedded computational resourcesChallenges - flexible network computing model, security and robustness
Vehicles with Sensors & Wireless
HealthcareApplication
Robotics Application
Network Connectivity& Computation
“Human in the Loop”
To Actuators
Virtualized physicalworld object
Content & LocationAware Routers
Computation& StorageProtocol
module
Ambient interfaces
Global Pervasive Network(Future Internet)
DataFromSensors
Smart Grids
“Cloud” Applications
MobilityFirst FIA Team Presentation Nov 15, 2010
Vision: Protocol Design for the future Mobile/Wireless World
Fundamental change in design goals and assumptions~10B+ mobile/wireless end-points as “first-class” Internet devicesMobility as the norm for end-points and access networksWireless access – varying link BW/quality, multiple radios, disconnectionsStronger security/trust requirements due to:
open radio mediumneed for dynamic trust association for mobile devices/users, increased privacy concerns (e.g. location tracking)greater potential for network failure
Mobile applications involve location/content/context and energy constraints
Technology has also changed a lot in the ~40 yrs since IP was designed
Moore’s law improvements in computing and storage (~5-6 orders-of-magnitude gain in cost performance since 1970)Edge/core disparity, fast fiber but continuing shortage of radio spectrum
MobilityFirst FIA Team Presentation Nov 15, 2010
Vision: Protocol Design for the future Mobile/Wireless World (cont.)
Proposed MobilityFirst architecture motivated by these considerations
Clean-slate protocol design that directly addresses the problems of mobility at scale, while also strengthening the trust model
End-point and network mobility at scaleIntrinsic properties of wireless mediumMore stringent security/trust requirementsSpecial needs of emerging mobile applications
Fixed internet access is treated as a special case of the more general designAlthough the “sweet spot” of our protocol is wireless/mobile, we believe that our design provides important benefits to fixed network applications
Security/trustRobustnessFault toleranceContext/content
Architecture Overview
MobilityFirst FIA Team Presentation Nov 15, 2010
Architecture: MobilityFirst Network Overview
Base Station
Wireless Router
AP
Core Network(flat label routing)
Router
Global Name Resolution Service
Control & Management Plane
Computing BladeBuffer StorageForwarding Engine
MobilityFirstRouter withIntegrated
Computing & Storage
Hop-by-HopTransport
GDTN Routing
Name <-> PKI address mapping
Data blockDataPlane
MobilityFirst key protocol features:
Fast global naming serviceSelf-certifying public key namesDynamic mapping of name to topological network address(es)Routers support both flat name and hierarchical address routingStorage-aware (generalized DTN) routing in accessHop-by-hop (segmented) transportProgrammable computing layerSupport for content/context/locationSeparate network mgmt plane
New components, very distinct from IP, intended to achieve key mobility and trust goals
More detailed rationale for the architecture in the following slides
MobilityFirst FIA Team Presentation Nov 15, 2010
Architecture: MobilityFirst Protocol StackCore elements of protocol stack (“narrow waist)
Global Name Resolution & Routing ServicesStorage-aware routing (GDTN) protocolHop-by-hop (segmented) transportServices and management API’s
Multiple TP and link layer options + programmable services
Link Layer A(e.g fiber)
Link Layer B(e.g LTE)
Link Layer C(e.g WiFi)
Link Layer D(e.g Ethernet
ManagementProtocol
Storage Aware Routing (GDTN)
Hop-by-Hop In-Network Transport
E2ETransport Protocol 1
E2ETransport Protocol 2
M-Socket
D-Socket D-Socket
M-APP
APP-1 APP-n
“Narrow Waist”
Global NameResolution
Service
NetworkService A
(e.g. Privacy)
NetworkService B
(e.g. Location)….
Network services (socket calls)Include:- send (name, data)
- Send(name1..name n, data)- Get (content_name)-- send/get (context_ID, data)-- send (location, data)
GlobalRoutingService
Programmable Services API
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Name-Address Separation
Separation of names (ID) from network addressesGlobally unique name for each network end-point
User name, device ID, content, context, AS name, and so onMultiple domain-specific naming services
Fast resolution & update from name address(es)Alternative designs possible
Routing on flat name spaceFlat name space with indirection to hierarchical topological addresses
Globally Unique Flat Identifier (GID)
John’s _laptop_1
Sue’s_mobile_2
Server_1234
Sensor@XYZ
Media File_ABC
Host NamingService
Network
SensorNamingService
ContentNamingService
Fast GID-Address Resolution Service
Network addressNet1.local_ID
Net2.local_ID
ContextNamingService
Taxis in NB
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Global Name Resolution Service
Fast Global Name Resolution a central feature of architectureDistributed service hosted by routers
Multiple competing providers to avoid single root of trustFast updates ~50-100 ms to support dynamic mobility (..home agent optional)Service will scale to ~10B names via P2P/DHT techniques, Moore’s law
AP
Global Name-Address Resolution Service A
Data Plane
Service B(competing providers)
Host Name
Network – NA2
Network – NA1
Network Address 1
Registrationupdate
Mobile nodetrajectory
Initialregistration
Name, Context_ID or Content_ID Network Address
Host Name
Network Address 2
Decentralixed name servicesHosted by subset of ~100,000+ Gatway routers in network
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Flat Name Routing Option
One option under consideration is a flat network identifier with a locally unique host identifier network address = net_ID.local_IDInterdomain routing based on flat network identifierSecurity through use of PKI-based self-certifying addresses (..similar to AIP)Some weaknesses: limited routing aggregation, no multi-homing, etc.
GID/ServiceHeader
Data Object
Name-GID Mapping
Flat Name Routing TableName GID
server@winlab Net123.localxyzDest GID Path
Net123 Net1, net2, ..
*MF Protocol Data Unit (PDU)for flat GID routing
GID/ServiceHeader
Data Object
Data Object
*Note1: MobilityFirst PDU contains the entire data objectranging from short message to large media file ~ GB
with hop-by-hop transport and storage (explained later)
*Note1: MobilityFirst PDU also contains a service definition fieldthat makes the PDU self-defining at each router in network. Thegeneral philosophy is to minimize per-packet/session/flow state at routers, and make basic services stateless.
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Hybrid Name/Address Routing Option
Another option under consideration is a hybrid scheme with support for both name (GID) and topological address (NA) routingNA header used for “fast” path, with fallback to GID resolution as neededAdditional level of indirection helps with router aggregation and supports more general name to address bindings e.g. dual-homing, content, geocast..
GID/ServiceHeader
Data Object
GID-Address Mapping
Routing Table
GID NA
12345.. xxyy, xxzz
GID/ServceHeader
Data Object
NA Header
Dest NA Path
xxyy Net1, net2, ..
Flat GID Routing(slow path)
Topological Address Routing(fast path)
Name-GID Mapping
Name GID
server@winlab Net123.localxyz GID/ServiceHeader
Data File
NA Header
Data Object PDU with name PDU with nameand address
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Storage Aware RoutingStorage aware (generalized DTN) routing exploits in-network storage to deal with varying link quality and disconnectionRouting algorithm adapts seamlessly adapts from switching (good path) to store-and-forward (poor link BW/disconnected)Storage has benefits for wired networks as well..
Storage Router
Low BW cellular link
MobileDevicetrajectory
High BW WiFi link
TemporaryStorage atRouter
Initial Routing Path
Re-routed pathFor delivery
Sample CNF routing result
PDU
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Segmented TransportSegment-by-segment transport between routers with storage, in contrast to end-to-end TCP used todayUnit of transport (PDU) is a content file or max size fragmentHop TP provides improved throughput for time-varying wireless links, and also helps deal with disconnectionsAlso supports content caching, location services, etc.
Storage Router
Low BW cellular link
Segmented (Hop-by-Hop TP)
Optical Router
(no storage)
PDU
Hop #1 Hop #2
Hop #3
Hop #4TemporarilyStored PDU
BS
GID/Service Hdr
Mux Hdr
Net Address Hdr
Data Frag 1 Data
Frag 2Data Frag n
……
Hop-by-HopTransport
More details ofTP layer fragmentswith addl mux hdr
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Context Aware Services
Context-aware network services supported at two layers by MF architecture
Dynamic mapping of arbitrary context or content label by global name serviceSame mechanism used to handle named contentOptional software services implemented at the computing layer
MobileDevicetrajectory
context_ID = geo-coordinates & Rutgers_student
Send (context_iD, data_pkt)
Context-basedMulticast delivery
Dest_Name=Context_ID
Context Name Resolution service
xx yyaa bb
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Computing LayerProgrammable computing layer provides service flexibility and evolution/growth path
Routers include a virtual computing layer to support new network servicesPackets carry service tags and are directed to optional services where applicableProgramming API for service creation provided as integral part of architectureSupport for virtualization of services
...... ......
Packet forwarding/routing
Computing layer CPU Storage
Virtual ServiceProvider
ContentCaching
Privacyrouting
anony
anony
MobilityFirst FIA Team Presentation Nov 15, 2010
Protocol Design: Management Plane
Separate mgmt plane designed into MF architectureProvides interface for cross layer parameters, configuration, faults, etc.Useful for querying time-varying connectivity of mobile user/networkMechanism for detecting and controlling security problemsClient-assisted collection of management dataDecentralized implementation of mgmt services via computing layer
AP
Control & Management
PlaneNet Mgmt Interface
(performance queries,
configuration, faults, security, ..)
Data Plane
Data packets
Concluding Remarks
MobilityFirst FIA Meeting Presentation Nov 15, 2010
Research Challenges, Hard Problems:Fast global fast name resolution service for ~10-100B end-pointsUnified support for mobile devices, context, location, content,… as an integral capability of the networkScalable routing protocol for flat name identifiers, taking into account unique mobility requirements (multi-homing, anycast, multicast, disconnection, location-awareness, …)
Design of storage-aware routing algorithms which are robust to disconnection and bandwidth variation across wired & wireless netsTechniques for interdomain network aggregation applicable to name-based and storage-aware routing under consideration
Integrating security and privacy features into network services – self-certification, multiple roots of trust, Byzantine fault-tolerance, …Design of network management plane features to achieve a high level of transparency and accountabilityNetwork neutrality, economic feasibility, …Evaluation & validation at scale!
Website: winlab.mobiltyfirst.rutgers.edu
MobilityFirst FIA Meeting Presentation Nov 15, 2010
Concluding Remarks:The team is truly excited about the prospect of designing, building and deploying a comprehensive Internet architecture We believe our approach has a number of novel components and is unique in its focus on mobility and future pervasive computing needsSome of the key building blocks of the MobilityFirst architecture are:
Fast name resolution serviceFlat identifier (or hybrid) routing with self-certifying ID’sStorage-aware routing (GDTN) and hop-by-hop transportSupport for advanced context, content and mobility servicesProgrammable computing layer for future in-network services
The project is still at an early stage and the design is expected to evolve over the next 9-12 months – watch for future updates!Outcomes of the project include architectural evaluation, dissemination to related Internet & mobile network communities, laboratory prototyping and real-world evaluations using GENI
Website: winlab.mobiltyfirst.rutgers.edu
Backup Slides:Prototyping Plan
MobilityFirst FIA Meeting Presentation Nov 15, 2010
Prototyping Plan: Phased ApproachMultiple stages of prototyping leading up to a full MobilityFirst protocol deployment by year 3 of the project
Prototyping of key protocol components for validation and measurement Early proof-of-concept, small-scale lab prototypes for controlled experiments Multi-site network, medium scale prototype for internetwork experiments Deployment of the MobilityFirst protocol as a stable, long-running VN on GENI campus networks for application trials with real opt-in users and evaluation of usability, trust
IndividualPrototypingOf KeyComponents
Small-scaleLab prototypeFor controlledExperiments
Multi-site Medium scalePrototype forInternetworkingValidation
MobilityFirstProtocolDeployment onGENI NetworkFor ApplicationTrials
Significant leverage/technical risk reduction from earlier protocol research projects at collaborating institutions, e.g. CNF at Rutgers, Hop TP at UMass/Amherst, Wireless net mgmt – Wisconsin and UMich; Internetwork routing – UMich; Android platform- UMass/Lowell
Year 1 Year 1,2 Year 2 Year 3
MobilityFirst FIA Meeting Presentation Nov 15, 2010
TP‐1TP‐1 TP‐2TP‐2
Network Layer
LL‐1 LL‐2
ComputingLayer forServices
LL‐3
Forwarding &
StorageEngine
NameResolutionService
ContentStorage
RoutingProtocol
RouterMemory
API‐1 API‐2
Compute API
ManagementPlane
Prototyping Plan: Router Software Implementation
Core router software implemented on Linux OSPortable to various networking platforms including Open BS, WiFi, Click and OpenFlowDistributed implementation (Naming Service – UMass; Routing/TP – Rutgers; Management – Wisconsin; Compute Layer – Duke) with integration at Rutgers
Click orOpenFlowRouter
Open BaseStation
Open AP
MobilityFirst FIA Meeting Presentation Nov 15, 2010
Prototyping Plan: Client Software
Linux and Android mobile implementation for clientsApplication repository including mobile social networking, content delivery and context-aware messaging
WiMax + WiFi Android Client
Also Linux PC/laptopClients with WiMAX and WiFI
MobilityFirst FIA Meeting Presentation Nov 15, 2010
Prototyping Plan: Experimental Platforms & Testbeds – ORBIT/GENI Testbed at Rutgers
ORBIT radio grid for reproducible evaluation at scaleOutdoor wireless network with open WiMAX BS, open AP and vehicular nodes for real world mobility experimentsTestbed connected to GENI backbone via L2 fiber (I2 PoP at Phila)
WiMAXOpen BS
VehicularNode
ORBIT Radio Grid ORBIT Radio Grid
MobilityFirst FIA Meeting Presentation Nov 15, 2010
Prototyping Plan: GENI/OpenFlow Campus Network at Rutgers
Outdoor ORBIT
GENI Open Base Station
(WiMax)Outdoor RU
Wireless
3.) Rutgers Outdoor Network
To MEGPI PoP(Philadelphia)
Rutgers Core Network
L3 -> L2
2.) ORBIT Grid
L2
Cook Campus
Busch Campus
AR
L2
OpenFlowEnabled Switch
OpenFlowEnabled Router
ARAccess Router(not OF)
LEGEND:
1.) SB9Outdoor ORBIT
GENI Open Base Station
(WiMax)Outdoor RU
Wireless
3.) Rutgers Outdoor Network
NetFPGA
NetFPGA
IP8800 IP8800
IP8800IP8800
GENI campus network with WiFi, WiMAX and OpenFlow to be used forinitial system level evaluations at Rutgers
MobilityFirst FIA Meeting Presentation Nov 15, 2010
Prototyping Plan: Outdoor Wireless Networks at UMass and Wisconsin
Madison Metro Transit city-buses thatinteract with WiFi mesh, WiMAX, and
3G cellular (60 sq. miles)
GPRS1
WLAN
EvDOWiMAX
WiFi
Mt. Toby
MA1
CSB
10.6 Km
Topology of the MA OTG testbed
Weather Observatory NetworkAt UMass
Dieselnet TestbedAt UMass
Additional campus networks at UMass, Wisconsin to be used for multi-site evaluation during second phase of prototyping
UMass and Wisconsin equipped with both WiMAX and WiFi (GENI Spiral 2)
Additional GENI campus site with WiMax planned for UMich under GENI Spiral 3
MobilityFirst FIA Meeting Presentation Nov 15, 2010
GENI Spiral 1 connectivity (figure courtesy of GPO at BBN Technologies)
Wisconsin
UMass
Rutgers
DukeTo GENI/I2Backbone
Open WiMAX/3GBase Station
End-user Android mobilePhones or Linux netbooks
with WiMax and WiFi
CampusTestbed
OpenFlow+ Blade ServerRouter Platform
Typical campusNetwork
MFA ClientSoftware
MFA Router Software
Campus sites
Prototyping Plan: Experimental Platforms & Testbeds – Multi-site GENI NetworkMulti-site GENI campus networks used for real-world evaluation of MF protocol during final stage of evaluation in year 3Involves realistic applications and opt-in users with both Linux PC’s or Android mobile devicesMF protocol as a long-running “slice” on GENI network