Paper #: [3.AppServer] 8
Title: Speculative Data Dissemination
and Service to Reduce Server Load,
Network Traffic and Service Time in
Distributed Information Systems


PROBLEM

Current research focuses on caching
improvements so that documents that are
accessed frequently may be saved on the
client's computer in order to
improve response time and limit the
work required by the server.  However,
a
more proactive approach is needed that
utilizes the data available on the
server. This approach needs to focus
not only on what the user has accessed
in the past, but more importantly what
is the user likely to ask for in the
future, which may or may not consist of
documents previously requested.

NEW IDEAS AND STRENGTHS

The authors propose two main ways to
improve data dissemination by reducing
network traffic.  The first way
consists of using services that take
into
consideration time and location.  These
services will receive information
from servers and distribute it to other
services in a tree structure that
are likely to utilize the information.

The other way utilizes speculative
services that look at the document
requested and sends along with that
document other documents that the user
may want in the future.  For example,
files embedded in the document would
definitely be sent along with the
document as well as others that are
related. The authors did a nice
analysis to determine that the benefits
of speculation are best when the
documents sent along with the requested
document are small.  This means that
even if the document is not used then
the wasted bandwidth is minimal.  On
the other hand, if the speculation was
correct then the savings in time and
bandwidth give performance improvements
far better than those attainable by
performing caching at clients.

WEAKNESSES AND EXTENSIONS

The paper brought up a good point with
the benefits of cooperative caching,
which entails the client sending a list
of document IDs for documents that
the client already has stored in his
cache.  This will prevent the server
from sending speculative documents
which the client already has.  However,
I
would suggest also sending timestamps
with the document IDs so that the
server may take into consideration
whether or not an update has been done
when determining which pages to send to
the client.