That paper is a nice way to get rid of downstream false patent claims from
others, protecting their investment. Good for them.

As stated, the Google system is heirarchical and closed-loop and they
already had a top-down architecture and a closed environment.

DHT for distributed 'service farms' (lets be accurate as it is mesh of
services) are better for environments that are more ad-hoc and open.
Amazon's Mechanical Turk, for example, has many joiners and leavers offering
services, which is not a top-down type of regime.


On 10/4/07, Ryan Barrett <[EMAIL PROTECTED]> wrote:

> On Wed, 3 Oct 2007, [EMAIL PROTECTED] wrote:
>
> > Yeah, I think you're late. I've heard repeated stories about google
> using
> > DHTs internally.
>
> i'm not aware of any of our infrastructure that uses DHT techniques at
> scale.
> we definitely have some sophisticated distributed systems, like chubby,
> bigtable, mapreduce, and our cluster management system, but they're all
> centralized or hierarchical, with master election.
>
> there may be some 20% projects, of course, but i don't see us using DHTs
> internally in the near future.
>
> (i'm not an official google spokesperson, look both ways before crossing
> the
> street, etc. :P)
>
> -Ryan
>



On 10/3/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
>
> Yeah, I think you're late.  I've heard repeated stories about google using
> DHTs
> internally.
>
> Thanks
> -greg
>
> Quoting David Barrett <[EMAIL PROTECTED]>:
> > So I've long been skeptical of the value of the DHT in the real world,
> often
> > arguing that -- unless anonymity and privacy are critical requirements
> --
> > central servers are far superior in every single way.
> >
> > However, I wonder if the real use for the DHT will be in vast server
> farms,
> > as suggested by Amazon's Dynamo project:
> >
> > http://www.allthingsdistributed.com/2007/10/amazons_dynamo.html
> >
> > I think it starts to make sense there, as we're talking about really
> > enormous scales were central servers truly do start to break down, where
> > churn operates on a geological timeframe compared to between clients,
> where
> > there are better trust guarantees and fairness is less a problem, etc.
> >
> > Is this common knowledge that I'm just late to pick up on?  Any
> thoughts?
> >
> > -david
> >
> > _______________________________________________
> > p2p-hackers mailing list
> > [email protected]
> > http://lists.zooko.com/mailman/listinfo/p2p-hackers
> >
>
>
>
> _______________________________________________
> p2p-hackers mailing list
> [email protected]
> http://lists.zooko.com/mailman/listinfo/p2p-hackers
>



-- 
Michael Slavitch
Ottawa Ontario Canada
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers

Reply via email to