CSC2231 - NOW review

A Case for NOW (Networks of Workstations)
-----------------------------------------
Thomas E. Anderson, David E. Culler, David A. Patterson & the NOW team

This paper argues the future importance of Networks of Workstations in
various applications from those needing a dedicated workstation to those
needing dedicated systems. After giving an analysis of supercomputers with
pros (communication performance, global system view) and cons (cost, time,
OS, application domains covered) the paper presents the belief and an
overview of the NOW project.

One of the key strengths of the paper lies in NOW's organization of
resources and its opportunities (memory - network RAM & Cooperative File
Caching, Redundant Arrays of Workstation Disks, Parallel Computing).
The project aproaches several issues such as low overhead, low latency
communication, the quality of the interconnections. Other aspect of the
project is its removing of single points of failure (e.g. xFS: Serverless
Network File Service, RAID implemented in software). Another interesting
approach is the construction of GLUnix from existing operating systems,
basically not reinventing the wheel when there is no need to do that.

It would be interesting to see a related work section (from that time).
Section 2.3 mentions the fact that using idle resources is not a new idea,
so I think that the related work on clusters of workstations should be
presented in more detail in a separate section.

I agree that the security should not be the real issue here, but the
good performance for all applications should. Being that this is more of
an overview paper I think that it should be followed by papers describing
each part of NOW's structure and opportunities (and it actually was).
Received on Wed Sep 14 2005 - 19:08:58 EDT