Rack-aware clients: save inter-rack bandwith when they select copies in the same cluster


If clients are aware of racks, and they select the copies in the same cluster, the inter rack bandwidth will be saved.


Acknowledge feature request.


Would extend this to detection of a local AS instance and usage of unix domain sockets instead of tcp loopback connections, if possible (assuming this will improve performance, atleast Redis is claiming to gain from it)

I am aware that you don’t recommend AS to run on same host as the app, but this setup will be a common one for small deployments.


It seems a better idea. :smile:


@ManuelSchmidt and @xiaosuo,

Thanks! We’ve noted this additional feature request.