Indexing234
Sure. Exactly the performance you get depends on a lot of things, but the
first few things to try are to use the _bulk API to load your documents and
to set the refresh interval to -1. To optimize the size of the bulk request
measure loading speed at a couple of mb per bulk and then increase it and
remeasure. Stop when it doesn't improve speed anymore or when you get to 20
or 30 mb. Those are fairly large bulks.
SSDs are much faster to write to, don't be afraid to use raid 0 and let
elasticsearch handle the redundancy. It is fairly normal to load with 0
replicas and then add them after load. It depends on your tolerance for
failure.
1 Like