2. 1 NINJA: A Service Architecture for Internet-Scale Systems Randy H. Katz UMC Distinguished Professor and Chair, EECS Department University of California, Berkeley Berkeley, CA 94720-1776 randy@cs.Berkeley.edu Ninjutsu is a stealth and espionage-oriented art which saw its greatest development in the 13th to early 17th centuries in Japan. Its practitioners, the ninja, were warrior-assassin-spies; most belonged to the Iga and Koga mountain clans. They were the supreme reconnaissance experts and saboteurs of their day.

3. 2 Presentation Outline Internet-Scale Systems A New Service Architecture Integration of Computing and Communications Opportunity for Cooperation Summary and Conclusion

4. 3 Presentation Outline Internet-Scale Systems A New Service Architecture Integration of Computing and Communications Opportunity for Cooperation Summary and Conclusion

5. 4 Internet-Scale Systems Extremely large, complex, distributed, heterogeneous, with continuous and rapid introduction of new technologies Feasible architectures –Decentralized, scalable algorithms –Dynamically deployed agents where they are needed; “Big infrastructure, small clients” –Incremental processing/communications growth –Careful violation of traditional layering Implementation approach based on incremental prototyping, deployment, evaluation, experimentation

6. 5 Emerging Communications Infrastructure of the Future The Challenge –Network-based applications becoming increasingly service intensive –Computational resources embedded in the switching fabric –Dealing with heterogeneity, true utility functionality, security & service discovery, in an open, extensible network environment Computing –Legacy servers –Partition functionality for “small” clients Communications –High bandwidth backbones plus diverse access networks »Third generation cellular systems »Home networking

7. 6 “From POTS to PANS: Telecommunications in Transition” “Today, the telecommunications sector is beginning to reshape itself, from a vertically to a horizontally structured industry. … [I]t used to be that new capabilities were driven primarily by the carriers. Now, they are beginning to be driven by the users. … There’s a universe of people out there who have a much better idea than we do of what key applications are, so why not give those folks the opportunity to realize them. … The smarts have to be buried in the ‘middleware’ of the network, but that is going to change as more-capable user equipment is distributed throughout the network. When it does, the economics of this industry may also change.” George Heilmeier, Chairman Emeritus, Bellcore

8. 7 Motivation: Why Can’t We... Control the A/V equipment in this room Get driving directions over the phone Affect other’s calendars directly Screen our calls based on the caller Control things with the interface we want Manage/sort/search our e-mail/v-mail/fax/news intelligently Needed: Ubiquitous connectivity and intelligent processing supported by the network

9. 8 NINJA: A Service Architecture that Provides... Tinkertoy wide-area components Automatic discovery, composition, and use Powerful operators: clusters, databases, and agents Viable component economics: subscription, pay per use Supports great devices, sensors, actuators Connects everything: ubiquitous support for access and mobility

10. 9 NINJA Computing Platform Information Devices (~10 Billion) Connected Stationary Computers (~100 Million) Scalable Servers (~Million) Units Active routers Bases “Small Devices, Big Infrastructure”

11. 10 NINJA Computing Infrastructure Computing resources inside the routing topology, not just at the leaves Paths chosen for location of operators as much as for shortest # of hops Mobile code that specializes the services provided by servers Mobility, management of bottleneck links, “integration” services, service handoff Server Client Proxy Router Compute Node

12. 11 NINJA Builds on Berkeley’s Systems Building Expertise BARWAN Wireless Overlay Networks Scalable Proxies BARWAN Wireless Overlay Networks Scalable Proxies MASH Collaboration Applications Active Services MASH Collaboration Applications Active Services RTPGateway Service Discovery vic, vat, wb TranSend TACC Model Wireless Access MASH Toolkit Active Services Model NOW/Millennium Computing Platform NOW/Millennium Computing Platform NINJA Scalable, Secure Services Computation in the Network “Smart Spaces” as an app Event-Response Programmable Access NINJA Scalable, Secure Services Computation in the Network “Smart Spaces” as an app Event-Response Programmable Access

13. 12 High-tier Low-tier Satellite High Mobility Low Mobility Wide Area Regional Area Local Area BARWAN Wireless Overlays: Ubiquitous Connectivity Seamless mobility among local and wide-area wireless networks via vertical handoff

14. BARWAN Proxies: Scalable Support for Heterogeneous Clients Cache control To Internet SAN (high speed) Utility (10baseT) Coordination bus $ Cache partition... Datatype-specific distiller $ $ $ $ $ $ html gif jpg Front End Front End PTM Control Panel Control Panel NOW Cluster

15. 14 BARWAN: Support for Thin Clients Top Gun Wingman –World’s only graphical browser for PalmPilot –Greater than 10,000 users worldwide Top Gun MediaBoard –Combines MASH tools with proxy –Interoperates with laptops, workstations Superiority of proxy-based approach to enabling “thin” clients successfully demonstrated

16. 15 Presentation Outline Internet-Scale Systems A New Service Architecture Integration of Computing and Communications Opportunity for Cooperation Summary and Conclusion

17. Vertically Integrated Services vs. Component Services Vertically Integrated Services Fixed composition Static deployment No reuse Mixed presentation/data Fixed UI Worse pieces Component Services Dynamic composition Rapid deployment Reuse Data only UI defined dynamically -- based on device/connection Competition at every level

18. 17 NINJA Service Architecture: Basic Elements of the Model Units: –sensors/actuators –PDAs/SmartPhones –Laptops, PCs, NCs –heterogeneous Active Routers: –soft-state –basestations –localization –local mobility support Bases: –highly available –persistent state –databases –computing –agents –“home” base per user –“global” mobility support

19. 18 NINJA Active Infrastructure Units: Client Devices Sensors & Actuators Active Routers: Active network routers Soft state Interchangeable Bases: Scalable, available servers Persistent state Service discovery Public-key infrastructure Databases Home Base User state E-mail User tracking “Smart Spaces”

20. 19 Example: Thin Client Access to Legacy Server Base AR Unit AR Sensor PDA Sensor Base Unit AR Legacy Server Home Base Personalization

21. 20 Example: Sensor Aggregation Base AR Unit AR Sensor PDA Sensor Base Unit AR Legacy Server Agent Aggregator

22. 21 Example: TopGun WingMan/Mediaboard Operation AR Mic PDA Camera Base PC AR Legacy Server Aggregator Multicast Connector MediaBoard Image Converter Un-Zip PDA Proxy

23. 22 NINJA Service Architecture Operators/Connectors/Interfaces Paths –Wide-Area Paths –Interface Interconnection –Path Optimization Services –Service Discovery –Automatic Path Generation Example Applications –Universal remote control/smart spaces –Universal In-Box/Personal Information Management

24. 23 Operators/Connectors Operators: –transformation –aggregation –agents Connectors: –abstract wires –ADUs –varying semantics –uni/multicast Interfaces: –strongly typed –language independent –set of AM handlers –Leverage all COM objects

25. 24 Wide-Area Paths Path is a first-class entity Explicit or automatic creation Can change dynamically –change path or operators Unit of authentication: delegate along the path Unit of resource allocation

26. 25 WAN Paths: Economics Key Idea: securely delegate authority to any path component –access your e-mail remotely –authorized services (subscriptions) –authenticated sensor data Digicash: –Pay-per-use services –Can be one-time (or subscription)

27. 26 WAN Paths: Optimization Insert operators into the path: –forward error-correction, e.g., for error-prone wireless links –compression/decompression, e.g., bandwidth constrained links Change parameters –retransmission windows –settings for wireless or satellites Reroute path dynamically

28. 27 Connecting Interfaces Goal: trivial to connect –can be automated Connectors are polymorphic –wires carry a generic ADU of type T –automatic marshall/unmarshall »code must be created at instantiation Enables automated connection and use

29. 28 Interoperability Wrapper operators for legacy servers –HotBot, Zip2, Patent server –Generic wrappers for each MIME type Connectors use TCP, UDP, multicast Leverage COM objects as operators –Control Excel remotely –Lots of third party components ODBC/JDBC databases

30. 29 Services Service –Highly available program with fixed interface at a fixed location Strongly Typed Interfaces –Multiple services of a given type compete –Compete on location, price, robustness, “quality”, brand name Service Discovery –Find “best” service of given type

31. 30 Automatic Path Creation Query goal is path creation Find logical path of operators –Path must type check Place operators on nodes –Some operators have affinity –Place them first Add connectors as needed Create any authentication keys

32. 31 Service Discovery and Use Four basic steps: Find a local Active Router Query Service-discovery service Automatic path creation, including transformations Generate UI from interface specification

33. 32 Application: Universal Remote Control Adapt device functionality to services in new environment –Beacon augmentation –Adaptive user interfaces –Composed behaviors Deployment within our building –Light, video, slide projector, VCR, audio receiver, camera, monitor, A/V switcher control –Local DNS/NTP/SMTP servers, HTTP proxies, RTP/multicast gateways –Audited printer access –Interactive floor maps, protocols for advertising object locations –Coarse-grained user tracking Universal Interaction?

34. 33 Generalization: Smart Spaces Walk into a A/V room and control everything with your own wireless PDA –Services for each device –Automated discovery and use –Automated UI generation –Composite behaviors –Local scope, no authentication (yet) Phones as well as PalmPilots –Speech-enabled control

35. 34 Experimental Testbed Network Infrastructure GSM BTS Millennium Cluster WLAN Pager IBM WorkPad CF788 MC-16 Motorola Pagewriter 2000 Text Speech Image/OCR 306 Soda 326 Soda “Colab” 405 Soda Ericsson Smart Spaces Personal Information Management Fax

36. 35 Target Applications: Personal Information Management Universal Inbox: e-mail, news, fax, voice mail Notification: e-mail, pager Priorities, Folders, Search Access: internet or telephone Peer-to-peer calendar Control of environment

37. 36 Personal Information Management Universal In-box Policy-based Location-based Activity-based Speech-to-Voice Mail Speech-to-Voice Attached-Email Call-to-Pager/Email Notification Email-to-Speech All compositions of the above!

38. 37 Presentation Outline Internet-Scale Systems A New Service Architecture Integration of Computing and Communications Opportunity for Cooperation Summary and Conclusion

39. 38 ICEBERG Vision How far can we go with a packet- switched cellular core network? How do you provision an IP network for large numbers of voice users? What new kinds of data-oriented services can be deployed?

40. 39 Mobility Management Mobile IP-GSM Mobility Interworking –Mobile IP-GSM authentication interworking –GSM HLR/VLR interaction in an Internet signaling context –Scalability of Mobile IP/hierarchical agents Multicast support for mobility –Alternative approach for mobility based on M/C addresses –Exploit multicast routing to reach mobile nodes without explicit handoff –Combine with real-time delivery of voice and video Generalized redirection agents –Policy-based redirection: e.g., 1-800 service, email to pagers, etc. –Redirection agents collocated with multicast tree branching points

41. 40 Packet Scheduling Validated VINT modeling suite for GSM media access, link layer, routing, and transport layers –GSM channel error models –Better understanding of sources of latency in cellular link and methods to circumvent this QoS-aware High Speed Circuit Switched Data (HSCSD), General Packet Radio System (GPRS), and Wideband CDMA (W-CDMA) link scheduling –RSVP signaling integration with bottleneck link scheduling –Fairness and utilization for TCP and RTP flows –Delay bound scheduling for R/T streams –Exploiting asymmetries in downstream/upstream slot assignment, CDMA self-interference

42. 41 New Services Proxies for Telephony-Computing Integration –GSM-vat-RTP interworking: handset-computer integration –Encapsulating complex data transformations »Speech-to-text, text-to-speech –Composition of services »Voice mail-to-email, email-to-voice mail –Location-aware information services »E.g., traffic reports –Multicast-enabled information services »Multilayered multicast: increasing level of detail as number of subscribed layers increase Demonstrate the speed and ease with which new services can be developed Develop innovative new services at the intersection of voice and data

43. 42 Potentially Any Network Service (PANS) 2-way Paging WIP GSM PSTN IP Iceberg Access Points (Beyond H.323 gateways) Provide policy engine Handle routing, security IAP Same service in different networks Service handoff between networks E.g., “follow me” service e.g., any-to-any service

44. 43 PANS Issues Entities: What are the endpoints? Naming: “What’s in a name?” Authentication: Entity to Entity authentication Billing: Charging entities, not “lines” Routing: QoS and cost issues Source conversion: Text-to-Speech, etc. Network management: Monitoring, provisioning Intelligent terminals: Services at the terminal

45. 44 OfficePSTN (Teaching): 510-642-8778 OfficePSTN (Chair): 510-642-0253 DeskIP: dreadnaught.cs.berkeley.edu:555 LaptopIP: polo.cs.berkeley.edu:555 PCS: 510-388-8778 Cellular: 510-409-6040 E-mail: randy@cs.berkeley.edu Home: 415-777-3382 OfficePSTN (Teaching): 510-642-8778 OfficePSTN (Chair): 510-642-0253 DeskIP: dreadnaught.cs.berkeley.edu:555 LaptopIP: polo.cs.berkeley.edu:555 PCS: 510-388-8778 Cellular: 510-409-6040 E-mail: randy@cs.berkeley.edu Home: 415-777-3382 “Randy@Berkeley” An Entity has a universal name and a profile; Entities are people or processes Universal Names: Globally unique IDs Profile: set of domain-specific names Service Mobility as a First-Class Object

46. 45 Iceberg Inter-Domain Naming Protocol Naming: –Reuse network’s local naming services –Single resolution point for universal names Routing: –Handles inter-network signaling –Users provide policies –IDNP gatekeepers provide policy engine –IDNP provides replication and consistency control

47. 46 IDNP Issues IDNP Server Call(Randy@Berkeley, Caller’s network, Interactive, CallerID certificate) IDNP Server Profile Policy System State Replicated Information: Real-time? Lazy? Epidemic? minutes/hours days/weeks weeks/months IAP

48. 47 Authentication and Billing Networks uses different mechanisms –PSTN autheticates “on-line” –GSM uses SIM card (Carrier-carrier agreement) »Original cellular networks used personal agreements –IP uses host address, X.509, etc. Iceberg certificates - two versions –Online verification by home network (Carrier) –Offline verification by local network (Personal/Visa)

49. 48 Call-Forwarding Example Called party can change policy and profiles from any network –Secure identification of entities –IDNP handles propagation of changes IDNP can expose domain-specific addresses or hide them for privacy Domain-specific information exposed to apps –Interactive? –Billing policies

50. 49 Presentation Outline Internet-Scale Systems A New Service Architecture Integration of Computing and Communications Opportunity for Cooperation Summary and Conclusion

51. 50 Berkeley Tradition of Experimental Computing Systems Research Evaluate existing technology to understand its weaknesses Deploy understand implementation complexities and sources of performance gain/loss Time Travel using today’s too expensive technology to prototype tomorrow’s systems Design new computing systems architectures

52. 51 Internet-Scale Systems Research Group Lead the evolution of the Internet through long-term research combined with the deployment of novel real- world large-scale systems and protocols –Unify on-going and future research projects in distributed computing, network protocols, services, access, new applications –Facilitate technology transfer and standardization –Work closely with industrial partners in an open laboratory environment We would like AT&T to be a charter member

53. 52 Benefits of Sponsorship Involvement with outstanding Berkeley graduate students Participation in large-scale, inter-disciplinary, pre- competitive research efforts with only modest investment, leveraging investment of other industrial partners Access to all ISRG-developed software, prototypes, simulation tools, and testbeds Early access to group’s research results through on-campus participation and retreats Support the expansion of cadre of researchers with expertise in Internet-scale systems

54. 53 Presentation Outline Internet-Scale Systems A New Service Architecture Integration of Computing and Communications Opportunity for Cooperation Summary and Conclusion

55. 54 Emerging Distributed System Architecture Spanning Processing and Access Computing and Communications Platform: Millennium/NOW Distributed Computing Services: NINJA Active Services Architecture MASH Media Processing Services Distributed Videoconferencing Room-scale Collaboration TranSend Extensible Proxy Services ICEBERG Computer-Telephony Services Speech and Location Aware Applications Personal Information Management and “Smart Spaces”

56. 55 NINJA and ICEBERG NINJA: Distributed Service Architecture –Service model based on Operators, Paths, Services –Platform model based on Units, Active Routers, Bases ICEBERG: Computer-Telephony Integration –IP-based backbone for cellular networks »Mobility and service interoperability in the context of diverse access networks »Performance issues: GPRS scheduling and IP scaling for mobile telephony applications »New services: Smart Spaces and PIM Internet Systems Research Group –“Lucent establishes research wing at Stanford” !?!?!?