Endeavour Retreat, June 19-20, 2000

ICEBERG Presentations/Posters

• Anthony Joseph, "An Update on ICEBERG"(ppt)
  Related Publication:
  
  • Helen J. Wang, Bhaskaran Raman, Chen-nee Chuah, Rahul Biswas, Ramakrishna Gummadi, Barbara Hohlt, Xia Hong, Emre Kiciman, Zhuoqing Mao, Jimmy S. Shih, Lakshminarayanan Subramanian, Ben Y. Zhao, Anthony D. Joseph, and Randy H. Katz, "ICEBERG: An Internet-core Network Architecture for Integrated Communications," IEEE Personal Communications (2000): Special Issue on IP-based Mobile Telecommunication Networks. (pdf)

Project Specific Breakout Sessions

• Ramakrishna Gummadi , Barbara Hohlt, "Using FSMs and Queues For Highly Concurrent Data Structures" (ppt) (ps)
  It is possible to achieve greater concurrency by combining a locking scheme with a scheduling scheme as compared to a separate lock and thread scheduling scheme. We have implemented a prototype single-node version of distributed B-trees, which are a class of Distributed Data Structures (DDS). We achieve scalability in handling queries within a single node by resorting to a finite state machine programming model, derive high concurrency in accesses by adopting a combined scheduling and concurrency-control (in particular, locking) discipline, and attain good performance while handling persistent objects by carefully controlling and optimizing buffer cache accesses. The entire B-tree architecture is designed to be non-blocking, and we generalize a few concurrency lessons learned from implementing B-trees using non-blocking FSMs and queues that we believe are universally applicable when programming in the FSM model.

• Jimmy Shih, "Congestion Pricing Design Review" (ppt)
  We propose using congestion pricing to allocate resources at access points. Examples of access points are modem pools for accessing the Internet, H.323 gateways for making and receiving telephone calls from computers, dorm routers for downloading music from Napster, departmental routers for sending and receiving IP packets. When using congestion pricing for each type of access network, we need to design a system to minimize unpredictability of prices and system behaviors. We plan to perform simulations to quantify benefits of using congestion pricing and deploy real systems to verify how effective pricing can be in affecting users' behaviors during times of congestion.

• Almudena Konrad, Amoolya Singh, "Experience with Video over GSM Wireless" (ppt, 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. To sending bit rate delay-sensitive traffic over a wireless link, suitable video compression algorithms and transport/link protocols are needed. We built 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. In the general case, this project provides a general wireless network infrastructure to support real-time streaming applications. We assume that the application layer provides error resilience, and packetization functionalities to the network. In turn, the network layers allow error-resilient policies at the application to be reflected in the transport and link layers by enabling flexible checksumming schemes such as UDP Lite and PPP Lite. We evaluated the overall performance of this wireless video system 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, and the Mean Opinion Square).

• Bhaskar Raman, "Walkthrough of the ICEBERG V0.0 Release" (ppt)
  ICEBERG Release Version 0.0 Components, Interfaces, APIs for extension, Limitations, Plans for next release

• Randy Katz, "ICEBERG Overview" (ppt)

• Sharad Agarwal, "Energy and Terrain Aware Heterogeneous Routing" (gif)
  This project looks at energy efficiency, terrain modeling and routing issues in ad-hoc networks of sensor nodes, gateway nodes and satellites.

• Chen-Nee Chuah, "Predictive End-to-End Reservations via a Hierarchical Clearing House" (ppt)
  Problem Motivation: Unpredictable packet loss and delay in IP networks can adversely impact the perceived quality of real-time applications like voice and video. Pre-partitioning or over-provisioning resources for real-time applications often results in inefficient network utilization.

  Approach: We have proposed to adapt reservations for aggregate real-time traffic based on a "Gaussian predictor", which uses on-line measurements of traffic statistics. The effectiveness of such approach is heavily dependent on the choice of measurement window. Our simulation study based on actual audio traces shows that dynamic reservations using one-minute Gaussian predictor can achieve average packet loss rate of less than 1% while incurring only 7% resource over-provisioning.

  Recent Progress: We extended the analysis to multiple domain case using a hierarcical Clearing House architecture to facilitate end-to-end aggregate reservations. In our model, we assume the network is composed of various network domains which can be aggregated to form a hierarchy of logical domains. A distributed CH architecture is associated with each logical domain to coordinate intra-domain reservations and perform admission control. We explore, with simulations, the efficiency of the CH architecture in terms of throughput, call blocking rate and reservation setup time.

  Next Steps: We plan to quantify the sensitivity of Gaussian predictor to both the measurement window and the time-scale of traffic changes, by experimenting with a vareity of workloads, including video and Internet traffic. We will evaluate different approaches to perform admission control, and explore how the Clearing House can leverage a network monitoring service.

  Recent Publication:

  • Chen-Nee Chuah, Lakshminarayanan Subramanian, Randy H. Katz and Anthony D. Joseph, "QoS Provisioning Using A Clearing House Architecture", International Workshop on Quality of Service(IWQoS), Pittsburgh, PA, pp. 115-124, June 2000. (ps.gz)

• Steven Czerwinski, "A General Infrastructure for Efficient Application-Level Protocols" (ppt)
  A design for an infrastructure of co-operating proxies in order to help reduce application protocol latencies through code and data migration.

• Ramkakrishna Gummadi and Barbara Hohlt, "Wide Area Events Using DDS" ( ppt)
  The Distributed Data Structre (DDS) API is a powerful abstraction on which to easily build cluster applications using reusable components. We show how the DDS can support the Publish-Subscribe-Notify class of internet services efficiently. As an example, the "Food in the Woz" application constructs a channel that notifies subscribers when food is available in the Woz lounge. The channel records user-preferences (such as the device that the user wishes to be contacted on) and delivers food events to subscribers after a transducer has recieved events from Ninjamail and a filter has generated events for each user.

  A paper on highly concurrent b-trees (ps)

• Steve Czerwinski , Barbara Hohlt , "Automatic Content Extraction for Voicemail Using Ninja", (ppt) (ps)
  Everyday, more and more devices are being hooked up to the global information network, wich partly consists of the telephony, cellular, Internet, and pager networks. There is a need for infrastructure support which effectively accesses information from heterogeneous networks and devices. Our approach is the use of transcoding operators for multi-modal information access and a mail access service which handles both regular email and voicemail encoded in MIME .  The Transcoder Service uses audio and text processing algorithms for extracting content from audio and text. Users may access all their messages from any mail respository in one place with calls to the Media Manager and have the content displayed on any device, be it a speaker or PDA. The Media Manager Mail Access Service handles mail related mime types along with our own mime attribute extensions, across any mail protocol. It uses the Transcoder service to perform transformations on mail content.

• Morley Mao, Wilson So, Frederick Wong, Shelley Zhuang, "Mobile Mote" (ppt)
  In many distributed sensor applications such as studying bird migration patterns, it is necessary to track down and retrieve data from a specific mobile object tagged with sensors and radio transceivers. Because these objects are mobile and operate over a wide area, it is expensive and sometimes impossible to build a single wide-area sensor network that covers all areas of operations. During period of connectivity, it is very important to retrieve information reliably from the individual sensors. In this paper, we present a system architecture that facilitates building such applications. Our proposed architecture consists of two components: a reliable link layer within each isolated local area sensor network, and IP connectivity between islands to support mobility in the wide area.

• Bhaskaran Raman and Z. Morley Mao, "Monitoring and Fail-Over in Wide-Area Service Composition" (ppt)

• Bhaskaran Raman, "Universal Inbox: Providing Personal Mobility and Service Mobility in an Integrated Network" (ppt)

• Jimmy Shih, "Using Prices to Allocate Resources at Access Points" (ppt)
  Problem Motivation: Access points are very congested. We plan to explore how prices can be used to allocate resources to those who value them the most during congestion. We also plan to explore how prices can be used to give users incentives to use the less capable access points.

  Approach: We plan to use the following two cases to explore using congestion pricing for admission control and load balancing. The first case is when lots of users want to use an Internet-to-PSTN gateway to make and receive telephone calls from their computers. The second case is when lots of users need to use an access router to send and receive IP packets. For these two cases, we plan to perform simulations and deploy actual systems.

  Recent Progress: During the fall of 1999, we had 12 CS graduate students using an Internet-to-PSTN gateway service to make and receive phone calls. During the spring of 2000, we had 50 students in the EECS department using our service.

  Next Steps: We plan to deploy our Internet-to-PSTN gateway service in the fall to 200 students in the UC Berkeley dorms. In the fall, we also plan to replace our research group's access router with a Nortel programmable router that uses prices to decide whose packets to drop during congestion.

• Lakshminarayanan Subramanian, "Aggregate Scheduling - Enhancing throughput in collective tasking systems" (ppt)

• Ben Zhao, "The Oceanstore Regenerative Wide-area Location System" (ppt, url)
  Service location has long been studied under a variety of assumptions and environments. Service location in the wide-area poses interesting problems which have not been completely solved. Some of these issues include true scalability in number of services, operation under network partitions and node failures, automatic recovery from data corruption, and protection from malicious users. In this ongoing work, we modify a recently developed wide-area data structure to address these issues, and present a service location mechanism well-suited for wide-area operation.

Last update: July 24, 2000
Contact the web-master, Chen-Nee