[Gluster-devel] AFR between two bricks over 3000 miles

Krishna Srinivas krishna at zresearch.com
Thu Apr 3 15:33:50 UTC 2008


On Thu, Apr 3, 2008 at 8:38 PM,  <nathan at robotics.net> wrote:
> On Thu, 3 Apr 2008, Krishna Srinivas wrote:
>
>
> >
> > >  Thanks! So is the ability to have two sites that are over 500 miles
> apart
> > > planed to be supported? I will test out the new code and let you know.
> > >
> >
> > Hi Nathan,
> > It is already supported but is slow as of now because of the slow link :)
> > But to make it fast over the slow link will be complicated, we have
> > not yet clearly
> > thought about how it can be achieved.
> >
>
>  That is just it, it is NOT a slow link. It is a full gig E between NYC and
> SJC. The issue is all the back and forth transactions. It does not matter if
> it is a T1 or a GIG E your only going to get about 100k miles a sec. So a
> request and then reply is going to be about 65 ms.
>
>  This is why people hack TCP/IP for sat links, you can have a full
> transponder to use, but you send your 1500 byte packet, 250 ms from earth to
> the sat back to earth and then the ack back another 250 ms. So you have this
> big pipe but you can't use it most of the time because of the delay.
>

Ah, OK. Code wise the change will be same to handle both the cases i.e
slow link and the delay due to distance. This should be handled in future,
but immediately we wont be able to fix it.

Krishna

>  Since gluster can use a number of flows at one time, you should be able to
> come close to maxing a gig e from east to west, but right now you can't even
> get close.
>
>
>
>  -Nathan
>
>
>  _______________________________________________
>  Gluster-devel mailing list
>  Gluster-devel at nongnu.org
>  http://lists.nongnu.org/mailman/listinfo/gluster-devel
>





More information about the Gluster-devel mailing list