If you have more indexed data than fits in the operating system page cache an HDD will generally perform much worse than SSDs, which is why the use of SSDs is recommended in the official performance guidelines.
Optimisations will depend on the nature and structure of the data, queries and access patterns, so it is hard to give general advice apart from what is available in the docs I linked to. It may help if you provided more details around this. As HDDs tend to have limited IOPS, limiting indexing (which can be IO intensive) to specific off-peak hours might be a good idea, although the tradeoff is that data may get stale. If you are able to go down this route it may be worth testing forcemerging shards down to a single segment once periodic updates have been made as that may reduce the IOPS required.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.