If space is not really an issue are there any other reasons to have the search factor lower that the replication factor? Thanks.
Hi @fredclown,
no, the only reason is storage saving because indexes are the most part of storage occupation (row data iare around 15% of the original data and indexes are around 35% or the original data).
if you haven't storage occupation problems, you can use the same SF and RF.
Ciao.
Giuseppe
This is really a border case but remember that indexes are built independently on each indexer so if you have SF=RF=8, you use CPU time for indexing on 8 indexers whereas if you have RF=8 but SF=2, you just stream the data to 6 indexers and store it and only on 2 indexers you prepare searchable indexed data for the bucket.
As I said, it's a border case since often the difference across your environment will be insignificant but there is a difference.
Also remember that only primary copy of the bucket participates in search so having a high SF lets you recover from disaster easier but doesn't speed up searches.
Hi @fredclown,
no, the only reason is storage saving because indexes are the most part of storage occupation (row data iare around 15% of the original data and indexes are around 35% or the original data).
if you haven't storage occupation problems, you can use the same SF and RF.
Ciao.
Giuseppe
Thanks sir.