[tahoe-dev] how to control the data distribution

[huang jun]

2012/4/18 David-Sarah Hopwood <david-sarah at jacaranda.org>

> On 17/04/12 13:36, huang jun wrote:
> > Hi,everyone
> > Tahoe using hash to balance the overload and data distribution cross the
> storage grid,we
> > have an  idea:
> > Assume the storage grid seperate in two sites A and B, there are gateway
> nodes in both
> > sites, and we want to store 4 shares of a file
> > on A, and 6 shares on B.so that when we read the file, we can direct the
> http request to
> > a nearest  gateway of our computer, take B as example,
> > then request the storage nodes in site B and get 3 of 6 shares, now we
> got the  file.
> >
> > so do we have any ways to achieve this?
>
> You can easily have multiple gateways to the same grid. Also, clients will
> automatically
> choose the nodes that respond fastest to download from, which are likely
> to be nodes on
> the nearest site in the situation you describe.
>
> What is not currently supported for this use case is to guarantee an even
> distribution
> of shares between sites. Also, if you have a mobile computer there is no
> way to
> automatically get it to connect to the nearest gateway.
>
>  thanks,
so if we want to achieve this, just selects the propotional servers in site
A and B may not work well if some nodes did not accept shares.
but to adjust the peer selection algorithm seems not a easy way.

> --
> David-Sarah Hopwood ⚥
>
>


-- 
huangjun
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://tahoe-lafs.org/pipermail/tahoe-dev/attachments/20120418/a99a94c3/attachment.html>