From: w...@tuhs.org (Warren Toomey)
Newsgroups: comp.os.research
Subject: Any work on SMP clustering?
Organization: Bond University
Approved: comp-os-resea...@ftp.cse.ucsc.edu
Originator: o...@cse.ucsc.edu
NNTP-Posting-Host: ftp.cse.ucsc.edu
Message-ID: <3d03a2a1$1@news.ucsc.edu>
Date: 9 Jun 2002 11:46:57 -0800
X-Trace: 9 Jun 2002 11:46:57 -0800, ftp.cse.ucsc.edu
Lines: 8
Path: archiver1.google.com!news1.google.com!newsfeed.stanford.edu!newsfeed.berkeley.edu!ucberkeley!news.ucsc.edu!osr
I came across Larry McVoy's idea: Scaling Linux with (partially) CC Clusters
(http://www.bitmover.com/cc-pitch/slide01.html) a while back. Does anybody
know of groups/individuals who are pursuing this idea? Contacts, pointers,
URLs much appreciated.
Warren
From: w...@tuhs.org (Warren Toomey)
Newsgroups: comp.os.research
Subject: Any work on SMP clustering?
Organization: Bond University
Approved: comp-os-resea...@ftp.cse.ucsc.edu
Originator: o...@cse.ucsc.edu
NNTP-Posting-Host: ftp.cse.ucsc.edu
Message-ID: <3d05081a$1@news.ucsc.edu>
Date: 10 Jun 2002 13:12:10 -0800
X-Trace: 10 Jun 2002 13:12:10 -0800, ftp.cse.ucsc.edu
Lines: 9
Path: archiver1.google.com!news1.google.com!newsfeed.stanford.edu!newsfeed.berkeley.edu!ucberkeley!news.ucsc.edu!osr
I came across Larry McVoy's idea: Scaling Linux with (partially) CC Clusters
(http://www.bitmover.com/cc-pitch/slide01.html) a while back. Does anybody
know of groups/individuals who are pursuing this idea? Contacts, pointers,
URLs much appreciated.
Warren
From: w...@henry.it.bond.edu.au (Warren Toomey)
Newsgroups: comp.os.research
Subject: Help finding SMP scalability papers
Organization: Bond University
Approved: comp-os-resea...@ftp.cse.ucsc.edu
Originator: o...@cse.ucsc.edu
NNTP-Posting-Host: ftp.cse.ucsc.edu
Message-ID: <3d3edcc8$1@news.ucsc.edu>
Date: 24 Jul 2002 09:58:48 -0800
X-Trace: 24 Jul 2002 09:58:48 -0800, ftp.cse.ucsc.edu
Lines: 39
Path: archiver1.google.com!news1.google.com!newsfeed.stanford.edu!newsfeed.berkeley.edu!ucberkeley!news.ucsc.edu!osr
I'm currently writing a technical report surveying the applicability
of microkernels to OS design, in particular to the problem of SMP
scaling.
I want to talk about the difficulties of constructing an OS that
scales nearly linearly on SMP systems. There are two aspects to
this problem. The first is the technical issue: ensuring that
processors don't stall (block, spinlock) while competing for OS
data structures, thus lowering the performance of the OS. The second
issue is that of the buildability of SMP OS, and how hard it is to
extend/modify/enhance/understand an OS where extensive fine-grained
locking has been used.
I'm having a hard time finding references to good papers that either
describe the problem in general terms, or describe the problems
faced in a real system. On the buildability side, I have:
* http://kerneltrap.org/node.php?id=222: A Larry McVoy interview
* http://www.bitmover.com/cc-pitch/slide01.html: Scaling Linux
with Cache Coherent Clusters
* http://www.bitmover.com/talks/cliq: Scaling Linux with SMP
Clusters
I could also dig up some Usenet/mailing list postings and web sites
on the SMP scaling issues faced by both Linux and FreeBSD. I still
need refereed papers: ones that give quantative measurements of
monolithic kernels on SMP, ones that identify the issues faced when
scaling kernels on SMP. I have read the K42 white papers (see
http://www.research.ibm.com/K42/). This looks like an excellent
solution, but I want papers describing the problem!
Many thanks in advance for suggested papers to read and review.
Warren Toomey
|