I found this paper difficult to read for several reasons.  One 
reason is that the authors made too many assumptions about the readers'
background in file system technology (at least in my case).  There was 
very little information in the introduction which discussed 
basic issues of parallel and distributed files.  Without a sound 
development of the basic issues, it was difficult to understand the 
discussions of the separate file systems.    

The other reason I had difficulty reading this paper was that it 
did not clearly explain why both distributed and parallel file systems
were discussed.  Do they have similar design issues?  How do their features
differ?  Are there different tradeoffs involved?  Although the summary
section does address some of these questions, it does not draw any 
conclusions draw any general conclusions about the similarities and 
differences of these two types of systems.

Suggestions:

The focus of this paper needs to be more clearly defined.  Are you 
attempting to show that most distributed and parallel file systems have
similar design issues?  

Explain *up front* why these issues (performance, scalability, security,
recovery, etc) are important in file systems.  In general, these issues
are present in lots of systems, and need to be clearly defined.

Eight systems is a lot.  I appreciated the thorough coverage, but some
measure of detail could have been sacrificed for brevity.  

A research paper needs to tell a story.  Try to spend more effort on 
leading up to the discussions of the file systems.  Provide an abstract;
briefly discuss basics of parallel and distributed file systems; provide
a strong conclusion which restates the thesis and major points. 

======================================================================

     *Well organized paper with good summaries for each of the 
      File Systems
     *Would like to have see more of the compare/contrast and 
      analysis of the various file systems, possibly suggesting 
      where we are headed. Perhaps the use of 5/6 papers instead
      of 8 would have allowed for a better analysis and comparison
     *Pictoral representations of the various FSystems may have
      helped better understand each of the File Systems - given
      the limited space to explain them in detail.

======================================================================

The authors present 8 distributed and parallel
file systems to shed some light on the diverse
design space of file systems.

I like the organization of the paper in the sense
that it provides a consistent interface to authors'
observations. However, the architecture and
implementation part of most systems are too vague to
draw conclusions. I guess the space limitation was a
big factor in this. Yet, the authors could have selected
one novel feature of each system and concentrated on
that to save space and provide a more detailed argument
on that feature.

I think the authors assumed too much about the 
knowledge of the target audience. Thus, certain 
points are not clear in the paper making it hard to
comprehend.

======================================================================

Import - My first reaction was "not another paper on file systems!", 
and it did not change substantially after reading the paper. 

Novelty - The summary of the eight arch's is nice and has been 
done thoroughly. But no new areas of future research possibilities 
or new problems have been highlighted/suggested in the paper.

Quality - The paper points out the disadvantages of the various 
arch's without many suggestions on how to overcome them. 

Overall - The paper may serve as a "summary" of the eight diverse
architectures and their comparison, but the paper does not 
elaborate on any particular aspect of file system design in depth to 
be of significant use.

======================================================================

This paper had a lot of good information, and the area itself is one
that is becoming increasingly important as gaps widen between CPU speed
and disk access time. 

The structure of this paper didn't resemble other academic papers that
we've read. I think it would have been better to introduce each system
briefly at the start and then divide the rest of the paper into comparison
sections: implementation, advantages, etc. This allows for more convenient
comparisons and doesn't force the reader to retain as much information to
"connect the issues." Most of the information presented felt disjoint, and
there was little done to tie it together.

The other major complaint that I had was that some information presented
was not backed with an example or further discussion. Many issues were 
just briefly touched upon without going into any substantial depth. For
example, the paper mentions that xFS has "more overhead and complexity in
locating files" and that it "requires cumbersome mechanisms for reaching
consensus," but there is no further clarification or support for either.

======================================================================