On Thu, Feb 23, 2017 at 12:21 PM, Ben Pfaff <[email protected]> wrote: > > On Thu, Feb 23, 2017 at 12:02:04PM -0800, Han Zhou wrote: > > On Fri, Feb 17, 2017 at 5:10 PM, Ben Pfaff <[email protected]> wrote: > > > > > + <p> > > > + Clusters larger than 5 servers will also work, with every 2 added > > > + servers allowing the cluster to tolerate 1 more failure, but > > > + performance decreases, especially write performance. > > > > Should it be clear that read performance should never be impacted? > > And is the performance drop supposed to be linear? If yes, then it might be > > justified to add more nodes to satisfy super large deployments whenever > > needed? Although I believe 5 nodes are good enough for most large > > deployments :) > > It's a little early to speculate on the performance of the > implementation, since it isn't implemented yet, but I'll try to add more > about performance.
Understood. > > There's also some question about the ideal mix between clustering and > replication. Combination of clustering and replication sounds good. So HVs can read from a more flexible number of replications, and under circumstances where occasional write is needed, HVs can use a separate channel to write to the cluster. Is this a valid direction to take? _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
