Browse Prior Art Database

Caching Pages for Reduced Network Traffic in Collaborative Sessions

IP.com Disclosure Number: IPCOM000116481D
Original Publication Date: 1995-Sep-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 2 page(s) / 70K

Publishing Venue

IBM

Related People

Chaiken, RI: AUTHOR [+2]

Abstract

By delivering only pages of immediate interest to users browsing a large document and leaving the perception that the whole document is available, the use of the network is optimal and efficient. The optimization is done over page size, which may vary from document to document, whereas file servers optimize over block size, which is a constant.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Caching Pages for Reduced Network Traffic in Collaborative Sessions

      By delivering only pages of immediate interest to users
browsing a large document and leaving the perception that the whole
document is available, the use of the network is optimal and
efficient.  The optimization is done over page size, which may vary
from document to document, whereas file servers optimize over block
size, which is a constant.

      The distribution of large files involves large overheads by way
of disk space, memory requirements and CPU times.  Since users can
view only page at a time, a system can provide enough pages to
satisfy users perusing a document at a reasonable rate without
overloading the users' disk or memory.

      The process of perusing documents is Poisson distributed.
Hence, pages in the immediate vicinity of the page being viewed have
the highest probability of being viewed next and pages farthest away
have the least probability.  But once a jump to a page at a remote
end of the document is made then the pages in the vicinity of this
new page have the highest probability of being perused.

      A user (client) joins the session by requesting a document from
the server.  If necessary, the document server rasterizes the
requested file.  The server returns N pages of the requested document
to the client machine.  If N is large, there is a delay in the data
getting to the client, and in the file handling activities once the
data gets to the client machine.  On the other hand, if N is small,
the user could receive too few pages.  Hence, N must be optimally
arrived at to satisfy both constraints.  N could also be computed
based the number of pages cached by the application.

      The first of N pages is displayed on the user's screen after
the readi...