Performance and cheap storage
brong at fastmail.fm
Tue Aug 8 01:59:52 EDT 2006
On Mon, Aug 07, 2006 at 10:47:10PM +0200, Phil Pennock wrote:
> On 2006-08-07 at 19:23 +0200, Hack Kampbjorn wrote:
> > Phil Pennock wrote:
> > >The "easy" fix is theoretically to configure up extra private addresses
> > >as aliases on the backend, and distribute the load over all of them.
> > >This avoids having multiple ports and multiple entries -- it's one
> > >cyrus.conf listening. The problem may be making sure that the front-end
> > >knows that several backends are tied together as being one real system,
> > >to avoid interesting failover effects.
> > Why not have the extra IPs on the front-end?
> Selecting the source address, where there are multiple source addresses,
> requires programming to manually force the non-default IP. The client
> needs to explicitly bind() a socket, then connect it to the remote host.
> If the software supports it, that's cool, but I was assuming that no
> programming was preferred. Sorry, it's the sysadmin in me.
> Adding the IPs on the backend involves no changes in programming.
Yes, exactly - though we're thinking about asking Igor (the author
of Nginx) to allow you to choose a local bind address for each
connection. We already send back everything else as HTTP response
headers from the authentication daemon. Using HTTP as an
authentication protocol felt strange for a bit, but it's actually
really flexible, and you can just run up a CGI for easy testing.
More information about the Info-cyrus