CS237: Spring 2011
Distributed Systems Middleware

List of Projects in the DSM Group
     
  Mobile Service / Grid Environments (6 projects). Contact Yun.  
     
 Topic 1:  Implementation of Data Placement Techniques for Grid-based MM Services ( 3-4 students )  
     
A grid system allows the use of idle computing and communication resources distributed in a wide-area environment. In this project, one would need to implement the data replication system for grid-based multimedia services using Globus Toolkits. We already have implemented the basic client/broker/server communication protocols, and the current system has a client' GUI, a broker-age scheduling module with network flow algorithm, a directory service in MySQL, and a multimedia player coded with JMF. So basically, in this project, we need one student to improve the scheduling algorithm, one student to enhance the directory service, and one student to design and implement the communication protocols:
  • A survey about grid storage techniques (e.g. work at IBM)
  • Design communication protocols between the broker (storage server) and grid volunteer servers
  • Implement the periodic placement and on-demand placement with updated information from directory service
  • A report of experimental results (e.g. the network transmission overhead and the replication delay)
  • A final demo

The detailed algorithms are represented in the conference paper [HiPC2002], please contact Yun Huang for the hardcopy.

 
 Topic 2:  Implementation of Machine-Status-Detector for Practical Study of P2P/Grid Environment ( 2-3 Students )  
     
  http://setiathome.ssl.berkeley.edu/ is the largest and most successful Internet distributed computing project to date. While millions of resource volunteers around the world run the program on their computers, the server does not have to know their current status. Unlike SETI@home, many grid systems make good use of Network Weather Service (NWS) to optimize their adaptive scheduling algorithms. So, this project is to implement a small module of the NWS, which detects the variations (e.g. CPU) of the volunteer servers in P2P or Grid environments and updates the data in an information center. By collecting the status information, the system can do a practical study about P2P/Grid environment using statistical learning theories. Basically, the project needs the following steps:
  • A survey about NWS and its applications in grid systems
  • Design the communication protocol between the information center and volunteer servers
  • Implement the Machine-Status-Detector
  • Install the Machine-Status-Detector on selective machines on campus, like ICS lab, graduates' offices, Libraries, students apartments... (we will support you and help you find machines)
  • Report experiment results (e.g. the CPU variations in a day and various behavior of machines at different locations )
  • Apply statistical analysis methods to study the P2P/Grid Environments and propose reasonable guesses

please contact Yun Huang to discuss the design and flow of the project, finding machine as well... :)

 

  
 Topic 3:  Proxy-based Remote Display vs. Online Video Streaming ( 1-2 Students)  
     
  Remote display techniques have been widely used to support multiple applications. Usually, there will be minimal system requirements to run a remote display program, such as the available network bandwidth of an end-to-end connection. This project aims to study the impact of using a remote display program on local machine and remote proxy. By comparing the cost (e.g. CPU, network...) of running a remote display software and providing an online video streaming service, we may identify what system conditions are more suitable for using remote display techniques.
  • A survey about remote display techniques (e.g. VNC) and systems (mmGrid of HP)
  • Design the experimental processes and choose proper software and applications
  • Report experimental results for above two services
  • Present the conclusion
  • A final demo

please contact Yun Huang for experiment details.

 

  
 Topic 4:  (Research Study) Information Management in Service Grid ( 1 Student )  
     
  In order to achieve various QoS performance and system scalability, a service grid needs efficient information management solutions. This project will apply different techniques used for fault tolerance in traditional distributed systems, popular Grid and P2P systems, so as to enrich the services' functional versatility and flexibility. Briefly, the project includes:
  • A survey about information managements in popular grids and P2P systems
  • Devise information management solutions for MM service Grid
  • Simulation of proposed approaches
  • Report simulation results
  • A final report

please contact Yun Huang for related research papers.

 

  
 Topic 5:  (Research Study) Agent-based Service Management for Wireless and Mobile Devices( 1 Student )  
     
  Offloading tasks from thin devices (wireless and mobile) onto other resource sufficient nodes is a promising and popular research topic now. Adapting to the current residual energy and moving patterns of a wireless and mobile device, a service scheduler can achieve power efficiency and better QoS performance. Imagine a situation that all service providers contact the same device at the same time with the same query frequently, this will put much overhead on the thin client with unnecessary cost. So, this project tries to provide an agent-based approach to managing a number of general services for each device. Since it keeps the current status of the mobile and wireless node, we name the agent as a "shadow" of the device.  Therefore, this project will include:
  • A survey about agent-based approach to providing services for wireless and mobile devices
  • Propose the "shadow" solutions
  • Simulation of presented ideas
  • Report simulation results
  • A final report

please contact Yun Huang for detailed research issues

 

  
 Topic 6:  (Research Study) Characterization of Mobility and Traffic Models for Wireless and Mobile Nodes( 1 Student )  
     
  Advances in wireless network have led to the advert of multiple innovative services, many (or most) of which need knowledge of mobile hosts' location, or should be able to predict the mobility patterns and traffic models to some extent. Quite a few people have been doing research related to mobility prediction, and their conclusions are of great benefit to future application design and system analysis, implementation, etc. In this project, you will do:
  • A survey about mobility prediction techniques and traffic model studies
  • Summarize a number of mobility patterns
  • Design a generator to simulate various mobility patterns
  • Implement the generator (prefer in Java)
  • A final report with demo

please contact Yun Huang for related papers

  
     
   
     
  Multimedia (6 Projects). Contact Hyunok.  
     
 Topic 1:  On-line Profiling Based Energy Reduction Video Conferencing  
     

This project aims to use the information available to the encoder to reduce energy consumed at the decoders. In a typical video conference, one device encodes the video stream and provides the compressed data to many decoding devices. If the encoder can provide decoding time to each decoder, then the decoder can apply DVS (dynamic voltage scaling) very efficiently. However, the problem with this is that it is very difficult to get the exact decoding time since the decoding devices have different platforms such as IPAQ, Palm, Zaurus, Axim and so on. Therefore, in this project you will figure out which information is necessary for each decoding device to predict decoding time and how one can collect the information in the encoding device.

The Project will involve

A Detailed Survey of related work in the area

·         Formulating the problem

·         Creating a new algorithm or method

·         Setting up the Experimental Environment

·         Experiment & Algorithm revision

·         A Final presentation

 
 Topic 2:  Quality Adjustment Considering Energy Budget  
     
 

In this project, you will need to work towards saving energy consumption in video conferencing. Let us take an example. Suppose there is an encoder, “A”, and 3 decoders, namely “B”, “C” and “D”. Assume that “D” has low energy capabilities. In order to reduce power consumption, lets say that “D” tries to skip P frames. However this skipping of P frames bears considerable image quality degradation if the ratio of I frame to P frame is large, like say 1:100, which means that 100 P frames follows one I frame. One possible approach to amortize this degradation into multiple decoders is that “A” changes the ratio of I to P frames to 1:10, which makes video quality worse for “B” and “C” but better for “D” since I frame requires more amount of bits than P frames.

You can investigate and come up with other approaches to address the problem.

The Project will involve

·         A Detailed Survey of related work in the area

·         Finding a problem

·         Creating a new algorithm or method

·         Setting up the Experimental Environment

·         Experiment & Algorithm revision

·         A Final presentation

 

  
 Topic 3:  Security in Video Conference  
     
 

To provide security requires additional computation and energy consumption. In this project, you will develop an efficient security algorithm which requires less computation but ensures security. One possible approach is to provide encryption for I frames instead of P frames since a P frame is always dependent on an I frame.

The Project will involve

·         A Detailed Survey of related work in the area

·         Finding a problem

·         Creating a new algorithm or method

·         Setting up the Experimental Environment

·         Experiment & Algorithm revision

·         A Final presentation

 

  
 Topic 4:  Trade-off between Security and Energy Consumption (or Computation)  
     
 

If you apply different security algorithms to a video encoder, you can get different computation time or power consumption. In this project, you will need to figure out the tradeoff between the level of security provided and the power consumed by varying security algorithms.

The Project will involve

·         A Detailed Survey of related work in the area

·         Finding and formulating the problem

·         Creating a new algorithm or method

·         Setting up the Experimental Environment

·         Experiment & Algorithm revision

·         A Final presentation

 

  
 Topic 5:  Multiple Qualities  
     
 

This project involves devising and implementing an encoding/decoding algorithm.

Let us consider an example. Suppose there is an encoder, “A”, and 3 decoders, namely “B”, “C” and “D”. Heterogeneous platforms of decoders have different display resolutions or computation power. This means that “A” should provide a different compressed image to each decoder. However, due to constraints such as computation or power capabilities and network bandwidth, it is not the most efficient approach. Instead, “A” should aim to provide one compressed data that has multiple qualities. Each decoder uses a part of the compressed data for decoding.

The Project will involve

·         A Detailed Survey of related work in the area

·         Finding a problem

·         Creating a new algorithm or method

·         Setting up the Experimental Environment

·         Experiment & Algorithm revision

·         A Final presentation

 

  
 Topic 6:  Video Conferencing System Specification by using PeaCE  
     
 

PeaCE is a software tool where one can specify an algorithm, simulate and generate implementing codes such as C and VHDL.

The video communication algorithm, which means 1:1 communication, has been implemented already. This project extends the previous implementation to support multiple user communication or video conferencing. The core of this project will be to design flow control.

 The Project will involve

·         Project specification

·         Implementation

·         Demo setup

·         A Final presentation

 

  
     
  Adaptive Middleware Framework / MPEG-4 (1 Project). Contact Vidya.  
     
 Topic 1:  Designing GURU - a MPEG-4 based middleware framework that can adapt multimedia content for users with disabilities. (3-5 students)  
     

In this project, you would need to design a middleware framework - GURU, that is modular and flexible. The primary task is to integrate existing adaptation modules to a fully functional adaptation framework. (additional modules to be designed if needed).  The middleware should be such that one can add more functional modules to it without having to modify the existing architecture. The basic adaptation framework is MPEG-4 based and  modules are designed using MPEG-J or Java. Details can be provided upon request.

The project will involve

  • A survey of multimedia or content adaptation frameworks.

  • Understanding the basics or MPEG-4 and MPEG-J.

  • Understanding of the goal of  GURU and its architecture which will be provided to you.

  • A detailed technical report, along with the implementation and documentation.

You will be provided with the modules that have already been developed. Documentation to the modules will also be provided. I will also provide the study material, guidance and help during the course of the project and there will be no time wasted in identifying the particulars of the project. I will also assist in getting details of the MPEG-4 standard, and also tools that might assist in the implementation process.
     
     
  Sensor Applications using GloMoSim simulator (2 Projects). Contact Qi.  
     
 Topic 1:  Querying in Sensor Networks (2 + 2 students)  
     
Sensors are typically deployed to gather data about the  physical world.  Since sensors are resource constrained, often sensor data is collected into a sensor database that resides at (more powerful) servers. A natural tradeoff exists between  the sensor resources (bandwidth, energy) consumed and the quality of data collected at the server. Blindly transmitting sensor updates at a fixed periodicity to the server results in a suboptimal solution due to the differences in stability of sensor values and due to the varying application needs that impose different quality requirements across sensors. Paper [HMV04] proposes adaptive data collection mechanisms for sensor environments that adjusts to these variations while at the same time optimizing the energy consumption of sensors.   
 
  • Project A. Quality-aware Aggregate Queries in Wireless Sensor Networks (2 students)
    • Paper [HMV04] only considers queries that request single sensor readings.  This project extends to aggregate queries which ask for the sum, average, min/max of a group of sensor readings. 
    • Hints: Explore query locality. subsequent aggregate queries may request data from a set of overlapping sensors, so obtaining an update from those sensors not only benefits current query, but also the next one.  In addition, sensors can coordinate with each other while deciding its state transitions.
    • Relevant papers: HMV04, OW00
       
  • Project B. Energy Efficient Data Collection in Wireless Sensor Networks (2 students)
    • There are basically three communication methods that sensor networks use: direct communication, multi-hop, or cluster-based. Paper [HMV04]  considers the scenario where direct communication between sensors and the server is more efficient than the other two manners. This project extends it to either multi-hop or cluster-based communication sensor networks (depending on your interests).  
    • Hints: Re-write sensor energy consumption formula in paper [HMV04] by taking into account energy consumption incurred due to the routing, and derive the optimal point where energy consumption is minimized.
    • Relevant papers:  HMV04, OW00, IGE00, HCB00
 Topic 2:  Fault Tolerant Data Management in Wireless Sensor Networks (2 students)  
     
  Sensor networks are subject to higher fault rates than traditional networks: connectivity between nodes can be lost due to environmental noise and obstacles; nodes may die due to power depletion, environmental changes or malicious destruction.  In order to guarantee certain level of delivery ratio of sensor readings,  a fault tolerant approach is desirable.

Hints:  This topic is very broad and has many possible directions to follow.  However, for this project, you are only expected to see how similar approach used for data management in mobile/disconnected environments can be applied in this scenario.  Why are they relevant?

Relevant papers:  KKGS01, TCO01