Currently the index is in memory only, so when the indexer is restarted, it must be built from scratch again.
It would be better to use an on disk index, and then use a lru ejection mechanism for the in-memory portion so that only hot items stay in memory and so we can configure the size of the in memory cache.
Currently the index is in memory only, so when the indexer is restarted, it must be built from scratch again.
It would be better to use an on disk index, and then use a lru ejection mechanism for the in-memory portion so that only hot items stay in memory and so we can configure the size of the in memory cache.