Large Scale Cluster using dsync and tagged director

[Matt Bryant]

Just been reading up on some of dovecot features and was wondering if
anyone had played around with large scale systems made up from mutliple
small dsync clusters and using tagged director to split users between
nodes of a sync'd cluster (rather than nfs).

I know traditionally its been shared storage and multiple nodes and
dovecots own doco suggest dsync for several thousands rather than
millions but given the cloud seems to be all about replication what
about many smaller nodes using dsync and scaling horizontally ?

The only issue is how does director scale when there are serveral(or
rather lots of) 'tags' ... in fact what are the limitations with director ?
 

Matt
-- 

<https://www.postbox-inc.com/?utm_source=email&utm_medium=siglink&utm_campaign=reach>