On Cache Prefetching Strategies For Integrated InfostationCellular Network - PowerPoint PPT Presentation

1 / 22
About This Presentation
Title:

On Cache Prefetching Strategies For Integrated InfostationCellular Network

Description:

On Cache Prefetching Strategies For Integrated Infostation-Cellular Network ... It alleviates bottleneck on the server, as well as saving the cost from ... – PowerPoint PPT presentation

Number of Views:68
Avg rating:3.0/5.0
Slides: 23
Provided by: unsw5
Category:

less

Transcript and Presenter's Notes

Title: On Cache Prefetching Strategies For Integrated InfostationCellular Network


1
On Cache Prefetching Strategies For Integrated
Infostation-Cellular Network
  • Jerry Chun-Ping Wang, Hossam Elgindy
  • School of Computer Science Engineering
  • University of New South Wales, Sydney, Australia
  • Justin Lipman
  • Intel Asia Pacific Research Development
  • Shanghai, China

2
Outline
  • Infostation Networks
  • Integrated Infostation-Cellular Network (IICN)
  • Cache prefetching problems
  • Our approach Longest Tree Distance Out (LTDO)
  • Results
  • Conclusion

3
What is Infostation
  • Infostation is a type of wireless network
    that combines many low cost, limited transmission
    range and high-bandwidth local wireless stations
    over an extended terrain. As mobile clients pass
    through the coverage area of an infostation, the
    information will be delivered to the clients. The
    information delivery process is aimed to be
    automatic and does not require any human
    interaction.
  • Types of information an Infostation might
    provide
  • - Location-dependent information
  • - Non-urgent information
  • - Distributive material
  • - Informative material

4
Features of Infostation
  • Advantages
  • Using combination of several discontinuous
    wireless LAN as oppose one large ubiquitous
    wireless WAN
  • Relatively cheaper equipment and access cost
  • Easy to setup
  • High data rate access for users in transiently
    close proximity to an infostation
  • Disadvantage
  • Non-ubiquitous coverage
  • Provides only opportunistic service
  • Clients have to access information in the absence
    of a network especially in between infostations.

5
Comparison between infostation and cellular
networks
6
Why integrating infostation and cellular network?
  • It became common to see devices to equip with
    multiple radios interfaces due to the
    commoditization of RF transceivers.
  • Utilizing multiple radios in an integrated
    manner to accomplish a common task would provide
    significant benefits in terms of functionality
    and performance.
  • 3G-WLAN inter-networking is also part of 3G
    system design.

P. Bahl, A. Adya, J. Padhye, and A. Wolman,
Reconsidering wireless systems with multiple
radios, ACM SIGCOMM Computer Communications
Reviews 34(5), 39-46,2004 3GPP. 3GPP system
to Wireless Local Area Network (WLAN)
Interworking Functional and architectural
definition (Release 6). 3GPP TR 23.934 V1.0.0.
7
Intergrated Infostation-Cellular Model
  • Information could be retrieved from two sources
    infostation and cellular networks.
  • Infostation continuously broadcasts data to its
    neighborhood in one direction only.
  • Cellular network provides data unicast via
    explicit requests from client.

8
Intergrated Infostation-Cellular Model (cont.)
  • The integration would effectively reduce the
    number of explicit requests due to broadcast
    advantage.
  • It alleviates bottleneck on the server, as well
    as saving the cost from accessing information via
    expensive cellular network.

9
Cache prefetching problems
  • We need a cache prefetching scheme to utilize
    broadcast well.
  • The caching management strategies such as LRU and
    LFU do not work well as they can not cache data
    prior to access.
  • FIFO blindly caching information in advance but
    will cause some important data to be thrown out.
  • PT calculates the product of access probability
    of item (P) and the time (T) that will elapse
    before that item appears on the broadcast, and
    store items with higher PT value.
  • PT calculate PT value on every clock tick, and
    require to obtain information from remote server.

S. Acharya, S. Zdonik and M. Franklin .
"Prefetching from a Broadcast Disk ". In
Proceedings of 12th International Conference on
Data Engineering (ICDE96), page 276, 1996
10
Requirements
  • We need a cache pre-fetching strategy that
  • utilizes broadcast by selectively caching data in
    advance.
  • uses client-oriented approach
  • does not require frequent calculation

11
Assumptions
  • Clients are browsing through a direct-access
    hierarchy of named catalogues and files from a
    remote server.
  • Each catalogue contains not only the relevant
    files to the catalogue, but also the
    sub-catalogues that contains refined information.
  • The file system forms a hierarchical data
    structure.
  • Every client starts the access from the root of
    tree.

12
Tree Distance
We define the Tree Distance (TD) as the
number of edges that is required to reach from
one item to another along the data hierarchy.
H(X) The size of set X A, B The set of
parents from the root to the selected item
13
Longest Tree Distance Out (LTDO)
  • LTDO evicts data item out of cache if the item
    has highest TD value.
  • LTDO is only based the clients access pattern
    and does not require to obtain the knowledge of
    broadcast queue in order to calculate the value.
  • It only calculates the TD value whenever the
    client is change its currently accessed item.
  • Cache can remain cohesive with the information
    hierarchy, thereby utilize the broadcast channel.

14
Simulation Environment
  • 4 infostations locates at each corner of a 3km x
    3km terrain, and each provides a non-overlapping
    coverage.
  • Each infostation has 6 clients (total of 24)
    within the neighborhood of infostation and each
    accessing information arbitrarily along the data
    hierarchy in a regular time interval.
  • Infostations broadcast data in sync according to
    the broadcast sequence provided by broadcast
    sequence generator.

15
Broadcast Sequence Generation
  • Hierarchical Partition (HPar) is a hierarchical
    formulation of a broadcast disk which is
    derived from index tree proposed by
    Imielinski.
  • Using Breadth First Search to traverse on data
    hierarchy and form a broadcast sequence.
  • Data Tree will be split into two smaller
    sub-trees. The sub-trees can be recursively
    split into finer sub-trees.
  • The degree of partition is defined as the number
    of recursion done on splitting the data tree.

T. Imielinski, S. Viswanathan, and B. R.
Badrinath. "Energy Efficient Indexing On Air". In
Proceedings of ACM SIGMOD Conference, pages 25--
36, 1994. Swarup Acharya, Rafael Alonso,
Michael Franklin, and Stanley Zdonik. "Broadcast
Disks Data Management for Asymmetric
Communication Environments ". In Proceedings of
ACM SIGMOD Conference, pages 199--210, 1995.
16
Broadcast Sequence Generation (cont.)
17
Simulation Results (FIFO)
18
Simulation Results (FIFO) cont.
19
Simulation Results (LTDO)
20
Simulation Results (LTDO) cont.
21
Conclusion
  • Introduced Integrated Infostation-Cellular
    Network Model
  • Infostation continously broadcast information to
    the clients in the neighborhood, while
    basestation provides unicast through explicit
    request.
  • Provided Longest Treed Distance Out (LTDO)
  • Evicts the data item that has longest tree
    distance.
  • The simulation has shown LTDO has great
    performance improvement than using FIFO.
  • Future Works
  • A better broadcast scheduling should be provided
    in accordance to client/server interaction.
  • Better calculation method for tree distance.

22
Thank you
Write a Comment
User Comments (0)
About PowerShow.com