Re: Fault tolerance background

[Tom Hobbs <tv...@googlemail.com>]

Hi Patricia,

I've gone through the same "Practical Byzantine Fault Tolerance" paper that
you have.  Here are some of my comments.  I don't if you'll find them useful
or not.

1.  "[replicatants] must start in the same state".  Does this mean that we
can't adhoc add or replace replicas?

2.  References 18 & 26 are probably worth checking out.

3.  We probably don't need the cryptography stuff from this algo.

4.  There is a limit to the number of replicants that must exist and be
working for the system to work.  That limit is > 2 which might be an issue.
I suppose it depends on what we say is required in any SLA.

Cheers,

Tom

On 24 Mar 2011 05:18, "Patricia Shanahan" <pa...@acm.org> wrote:
> I've started writing an annotated bibliography for background reading on
> fault tolerance for River developers. It is checked in to
> http://svn.apache.org/viewvc/river/jtsk/skunk/patsFaultTolerance. The
> check-in includes the generated .pdf file, so that anyone can read it
> even if they are not set up to process bibtex and latex files.
>
> If you read any papers in this area and want to add them, you have two
> choices:
>
> 1. Check out the directory, edit the .tex and .bib files, rebuild the
> .pdf, and check in the new versions. This option is only recommended if
> you are already set up to process latex and bibtex files.
>
> 2. E-mail me the citation and your comments and I'll incorporate them.
> If convenient, please include the bibtex citation.
>
> Note that even if a paper with an apparently relevant title turns out to
> be useless, it is still worth adding it with a note on why it is
> useless. That way, other developers will be able to avoid wasting time
> finding and reading it.
>
> Patricia