If I do this on a client that's solely dedicated to copying out and in again, will there be a performance hit on all the other clients using it? I really like the simplicity of glusterfs, hats off to the dev guys! > -----Original Message----- > From: Kirby Zhou [mailto:kirbyzhou at sohu-rd.com] > Sent: 15 July 2009 16:45 > To: Hiren Joshi; 'Vikas Gorur' > Cc: gluster-users at gluster.org > Subject: RE: DHT with AFR > > AFAIK, There is no method to redistribute your already stored > files except > copy out then copy in again. > > > > -----Original Message----- > From: Hiren Joshi [mailto:josh at moonfruit.com] > Sent: Wednesday, July 15, 2009 4:52 PM > To: Vikas Gorur > Cc: Kirby Zhou; gluster-users at gluster.org > Subject: RE: DHT with AFR > > > > > -----Original Message----- > > From: Vikas Gorur [mailto:vikas at gluster.com] > > Sent: 15 July 2009 09:01 > > To: Hiren Joshi > > Cc: Kirby Zhou; gluster-users at gluster.org > > Subject: Re: DHT with AFR > > > > > > ----- "Hiren Joshi" <josh at moonfruit.com> wrote: > > > > > My thinking is both on the client so: > > > I AFR my nodes. > > > I then DHT my AFR bricks. > > > I then mount the DHT vols. > > > > > > Or would I get better performance the other way around? > > > > DHT over AFR'd pairs is the configuration you want. You can > > then add another AFR pair whenever you want to scale up. > > When I add another pair, is there a way of re-distributing the data > evenly? Will this have a big performance hit? > > > > > Vikas > > -- > > Engineer - Gluster > > > > >
