January 2000 Retreat
[ Talks | Posters | Breakout Sessions | Handouts/Notes ]

• Talks
  • Opening Talk: ISRG Review, by Randy Katz
    (Powerpoint)
  • Touring ICEBERG, an Overview and a Tutorial, by Helen Wang
    Abstract:
    In ICEBERG, we are building an Internet-core network for integrated communications. Our network can be viewed as islands of cluster computing platforms that offer scalable, available, and robust services on them. Our signaling protocol takes the approach of lightweight sessions and soft state, enabling scalable and robust call services in the wide area. With multi-device communication as the basic call service combined with ICEBERG components that manage user preferences, behaviors, and compose units of computation, we provide an easy service creation environment for customizable services. To address wide area quality of service issues, we use a Clearing House-based approach that leverages aggregation to provide scalable resource reservation and billing mechanisms. The ICEBERG architecture also provides secure, authenticated call setup and billing for services. We have a significant amount of future work ahead of us, in particular we have not yet addressed the problem of Operations, Administrations and Maintenance (OA\&M), a critical technology that is mature in modern telecommunications networks. In addition, we are continuing our experiments with more sophisticated and novel services and we are in the process of formulating a well-defined service creation model. We also plan to address the incremental wide area deployment of ICEBERG architecture and Clearing House infrastructure over the existing Internet.
    (Powerpoint 97)
  • A Security Framework for a World of Post-PC Clients and Infrastructure-based Services, by Steve Ross
    Abstract:
    The Internet has made a tremendous variety of services and content available to users. At the same time, the Post-PC revolution has brought the demand for information access beyond the desktop to a wide range of new devices with differing capabilities, ranging from many small information appliances to public access terminals. We propose the use of an infrastructure based security framework to bridge the gap between the capabilities of Post-PC devices and security requirements of the services. The framework enables secure access from a wide variety of devices including computationally challenged PDAs as well as untrusted public terminals. By using generic content and security transformation functions to decouple device capabilities from service requirements support for new devices and services can be easily added to this framework. It is novel that this framework has the ability to deal with un-trusted public Internet access points by providing fine grain control over the content and functionality exposed to the end device. (Powerpoint)
  • Stepping Stones to Ubiquitous Connectivity, by Jason Hill (Powerpoint)
  • The ICEBERG H.323 Computer Telephony Service, by Jimmy Shih
    Abstract:
    In the ICEBERG (Internet CEllular BEyond the thiRd Generation) Project, we want to explore research issues associated with integrating computer and telephony technologies. To explore these issues, we have built a service that allows people to use their computers to make and receive telephone calls.
    We have a Motorola H.323 gateway that connects the Internet to the Public Switched Telephone Network (PSTN). We use Lucent's Elemedia H.323 protocol stack to build a proxy that sits infront of the gateway to perform admission control, accounting, routing. To use a computer to make an outgoing PSTN phone call, a user can use a H.323 client, like the Microsoft NetMeeting, to call the proxy with the number he wants to reach. The proxy will first perform admission control and accounting, then ask the gateway to route the call to the appropriate PSTN phone. To receive incoming PSTN phone calls on a computer, a user needs to first use his H.323 client to update his IP address to the proxy so that the proxy knows how to route that user's incoming phone calls. Then when someone on a PSTN phone wants to call that user on his computer, she will need to call the gateway and enter a four digits number for that user. The gateway then forwards the incoming phone call to the proxy to perform admission control, accounting, and routing.
    We are in the process of offering this service to the EECS community so that we can explore research issues like scaling, multiple service providers, congestion pricing, and multicast billing.
    (Powerpoint)
  • XSet: Lightweight XML Database, by Ben Zhao
    Abstract:
    As computing services and infrastructure move more and more to the distributed model, there is an increasing amount of metadata that needs to be indexed and searched as components inside applications. They had previously not been thought of as database applications, because of the heavy-weight nature of databases in performance and semantics. XSet is a small main memory XML database which offers performance and increased concurrency by exploiting the weak level of semantics required by these metadata applications. It has already been integrated into some Ninja applications, and have been used in a variety of applications outside of the Berkeley domain.
    (HTML)
  • Fault-Tolerant Paths, by Morley Mao
    Abstract:
    The path creation service has been newly designed to allow easy manual path creation and path maintenance. The goal of the service is to allow easiy, robust and fault-tolerant service composition. A new prototype has been built that lets users create paths manually and query the operators' running status. Automatic path patching algorithm has been designed that guarantees the fault-tolerance and robustness of running paths. The example application is the Ninja Jukebox service. This application allows users to listen to music from cellphone or vat. The path service builds a path consisting of transcoders changing from MP3 format to GSM format. If any operator and communication link between operator should fail, the path service automatically repairs the path by restarting the failing operator and creating another communication channel.
    The logical path matching algorithm has been refined. It is basically a shortest path graph search problem. Any well-known single source shortest path is applicable to this. The Dijkstra greedy algorithm has been implemented to achieve the running on the order of O(n^2) (n: number of nodes in the graph). To map path search problem to a graph problem, the type of input and output becomes nodes in the graph. An edge exists between two nodes, if there is an operator that does the transformation between the two types. To optimize path search problem for different criteria, the cost of the edge in the graph is modified to reflect the goals of optimization. XML is used to describe the properties and requirements of operators.
    (Powerpoint)
  • Universal In-box, by Bhaskaran Raman (Powerpoint 97)
    Elevator talk:
    How nice would it be if it were possible to have your pager messages read out to you on your cell-phone; based on who sent the page, where you are, what you're doing, and so on? How do you build it so that you can have the same service extended to a new cool device that operates on a BlueTooth interface? Can you have the right building blocks to have this extension implemented and deployed in a few days/weeks' effort?
    Abstract:
    In this presentation, we present our experience in building the Universal In-box in the Iceberg network. The Universal In-box refers to a set of capabilities that encapsulate (a) integration of communication mechanisms in different networks, and (b) personalization of the communication management features. We describe how the Iceberg architecture has allowed us to easily extend the Universal In-box to a new communication service: an instant-messaging system - this extension took minimal effort (less than a day for a no-frills implementation). It does this by abstracting away network and device dependent details in the middleware components and away from the one building the service. Although the architecture provides such extensibility, the component set, and the interfaces between the components may not yet be generic enough for other kinds of services. Another shortcoming has been in addressing the scaling concerns. We plan on looking at these issues in the near future.
  • Design of a Scalable Clearing House Architecture, by Lakmi Subramanian, Chen-Nee Chuah, and Ramki Gummadi
    Abstract:
    A Clearing House (CH) coordinates the interactions between the various Internet Service Providers (ISPs) in the network. The structure of the CH is designed for efficient provisioning of resource reservations and sercure billing services. In addition, the CH must scale over both distance and number of ISPs to support the operation of ICEBERG in the wide area. Three key ideas emplyed in this design to make the CH scalable to a large user base are hierarchy, aggregation, and indirection. We assume the network is composed of various basic domains (based on either administrative or geographic boundaries) which are aggregated to form logical domains This introduces a hierarchical tree of the logical domains and a ditributed CH architecture is associated with each logical domain in the gree. Each of these distributed architectures maintains only the aggregated state of their domains. Call setup time is reduced by employing powerful statistical techniques like Kalman filters to estimate the call traffic across various links and perform advance reservations. The report also discusses the scalable security, billing and privacy issues in the Clearing House.
    (Powerpoint)
  • Personal Activity Coordinator, by Xia Hong
    (Powerpoint)

• Posters
  • In Search of a Service Platform for ICEBERG, by Helen Wang
    Abstract:
    This poster intends to generate discussions on what direction we should take on defining a service platform for ICEBERG, namely a service creation environment (or SDK) for ICEBERG Services. ICEBERG services can be divided up into two categories: Internet service access from heterogeneous devices and customizable communication services. For the first type of service, we propose the use of a virtual IAP to make all Internet services behave like an ICEBERG endpoint. For the latter, we intend to define a complete set of observable events (customizable points) in communication services. To mitigate standardization, we are investigating some compiler techniques to enable the mix and match of service logics and underlying communication systems.
    (PowerPoint 97)
  • XSet, a lightweight XML database, by Ben Zhao
    Abstract:
    As computing services and infrastructure move more and more to the distributed model, there is an increasing amount of metadata that needs to be indexed and searched as components inside applications. They had previously not been thought of as database applications, because of the heavy-weight nature of databases in performance and semantics. XSet is a small main memory XML database which offers performance and increased concurrency by exploiting the weak level of semantics required by these metadata applications. It has already been integrated into some Ninja applications, and have been used in a variety of applications outside of the Berkeley domain.
    (HTML)
  • Resource Provisioning and Bandwidth Brokering for IP-core Networks, by Chen-nee Chuah
    Abstract:
    We describe the Clearing House approach to perform resource reservation and bandwidth brokering, which are critical for providing quality of service (QoS) in the wide-area networks. We discussed the traffic monitoring mechanisms involved, and algorithms we use to predict bandwidth usage in different time-scales. The main ideas behind our reservation strategies are: aggregation, hierarchy and decoupled signalings between different Clearing House nodes. Resource reservations are not performed on a per-call basis, but aggregated over various connections that use a particular link. The hierarchical Clearing House structure allows an efficient distribution of global information required for making admission control decisions and dynamic adjustment of resource reservations. The decoupling of notifications and reservation requests reduce the signaling overhead and reservation setup time.
    (Powerpoint)
  • Video Over Wireless Links, by Almudena Konrad and Amoolya Singh
    Abstract:
    The wireless link poses a significant challenge for sending video streams. This is due to the fact that current generation wireless links have low bit rate and high error rate compared to wire-line links. In order to send high bit rate delay-sensitive traffic over a wireless link, suitable video compression algorithms and transport/link protocols need to be implemented. Our goal was to build a wireless video system merging a low bit rate video codec with appropriate transport/link layer protocols that overcome the constraints imposed by the wireless link. The overall performance of this wireless video system was evaluated using both quantitative performance metrics (such as throughput, jitter, packet loss, and end-to-end latency) and qualitative performance metrics (such as viewer perception of quality: smooth motion, changes in luminance, edge detection, etc).
    (Powerpoint)
  • Internet Service Models by Ramki Gummadi
    Abstract:
    Existing Internet Service Models are inadequate because a) don't allow efficient service composition b) don't have an efficient event model that captures the context of use of a device c) have no unified way of dealing with persistent user preferences and data (such as billing, previously manipulated content, etc.), as well as ephemeral data (stock quotes, weather reports, traffic information, etc.). In the poster, we present a service architecture that facilitates dynamic service composition through active proxies, has a simple and powerful service naming, creation, migration, and usage mechanism, exposes an event specification and notification mechansim, and deals with transient and persistent data in a single framework.
    (Powerpoint)
  • Automatic Content Extraction for Voicemail Using Ninja by Steven Czerwinski and Barbara Hohlt
    (PowerPoint)
  • Service Creation and Extensibility in Iceberg , by Bhaskaran Raman (PowerPoint 97)
    Elevator talk:
    Suppose that you have a Jukebox service that can play mp3 music to your desktop on the Internet. What does it take to have the same service available on your new wireless hand-held device with audio capabilities? How much effort does it take for someone to such an extension? Can you have the right middleware components to have this done in a matter of days/weeks?
    Abstract:
    This poster explains the experience in extending two Ninja services to Iceberg devices. The two services were (a) a jukebox service that plays mp3s on IP-hosts, and (b) an instant-messaging system that integrates email, AIM, and ICQ. Both were built completely independent of Iceberg. Conceptually, as well as in implementation, extending these service to Iceberg end-points has been very easy. With the same amount of effort, the service is extended to ALL Iceberg end-points. This was possible due to the network and device independence features provided by Iceberg components.
  • Security Issues in Iceberg Call Setup , by Bhaskaran Raman (PowerPoint 97)
    Elevator talk:
    What are the authentication issues associated with personal mobility in Iceberg, where the components reside in the Internet? When you have a network component that redirects and filters your calls, how do you ensure that it is not by-passed? Can you have a capability whereby you can with-hold your cell-phone number from someone, and still have calls from them come on your cell-phone on special occasions? How do you do this in a secure fashion resilient to attacks?
    Abstract:
    This poster presents our initial thoughts on authentication mechanisms for secure call-setup in the Iceberg scenario. The mechanisms for 3 specific steps are detailed: (a) Naming service lookup, (b) Preference registry lookup, and (c) the actual call setup phase. We assume a PKI; the naming service can use the PKI for authentication, and can also serve as a location independent mechanism for distributing keys. The second step involves the issue of caller-id authentication; and the third involves that of ensuring callee control, and callee number privacy. Several optimizations are possible to avoid all of the authentication steps. We have not yet considered the steps involving other Iceberg components: the APC-Service, the Clearing-House, and the Personal-Activity-Coordinator.
  • Pricing, Charging, and Billing Experiments using the H.323 Gateway, by Jimmy Shih. (Powerpoint)
  • Iceberg Preferences Manager, by Rahul Biswas.
    Abstract:
    The Preference Manager is a graphical tool that allows users to specify their communication handling preferences for the Iceberg system. Communication includes incoming stream data such as telephony or fax and static data such as pages and e-mail messages. With this tool, the user can choose to route incoming communication to any end device based on the identity of the caller, the type of device the caller is using, the time of day, or dynamic state from the Personal Activity Coordinator. These preferences are used to generate a script that is then stored on the local Preference Registry. The Preference Manager features group aggregation, call waiting screening, an automated locator service, and a call simulator.
    (Powerpoint)

• Breakout session discussions
  • Event-model for Internet Services, reported by Ramki Gummadi
    Abstract:
    Current Internet Services have no generic mechansim of using events to provide context-dependent services. We studied a model for event subscription, publishing, posting, notification, and delivery that is both simple and powerful to be used directly in scenarios like persistent NinjaMail service, and SDS service. The model also supports filtering of events, chaining of filters, and service composition through delivery of events to other services, through the concept of a channel, filter, transducer, and sink.
    (Powerpoint)
  • Wide-area paths , reported by Ben Zhao
    Abstract:
    The issue of APC has been tackled recently by two separate projects. While there is much work to be done, the local area story seems to be tractable, and a complete implementation forthcoming. But what about the wide area? Already the ICEBERG model is pushing for APC functionality on the wide-area, and the problem still seems to be completely intractable. We had a relatively fruitful discussion on the importance of the problem, whether it was tractable, and came up with several best-effort approximations to a generalized solution.
    (Powerpoint)
  • Performance Metrics for ICEBERG, reported by Jimmy Shih
    Abstract:
    The mission of ICEBERG project is to build an Internet-core network architecture and service infrastructure that provide integrated telephony and data services across heterogenious access networks. Our primary goals include scalability, robustness, operation in the Wide-area, personal and service mobility, ease of service creation,security and privacy. The next challenge we encounter is to derive a set of metrics for evaluating the success of ICEBERG in achieving these goals.
    In particular, we are looking for feedback on:
    • the appropriate simulation framework or testing platforms to study the scalability performance of ICEBERG
    • reasonable metrics to measure ease of service creation (for developers), ease of usability and configuration (for users) of the various services provided by ICEBERG
    • well-studied benchmarks we can use to evaluate ICEBERG performance
    • privacy issues and vulnerability of our security model.
    Breakout Overview (Powerpoint)
    Breakout Results (Powerpoint)

• Handouts/Notes (if not direct from poster/talk)
  • Retreat Industrial Feedback, (HTML)
  • Design For a Clearing House Architecture, (Postscript)
  • Automatic Content Extraction for Voicemail, by Steven E. Czerwinski and Barbara Hohlt. (Postscript)
  • Feedback / Breakout on SS7 and Reiner's Challenge, (Text)