In-memory document storage in Vespa.ai

Question

Is there any way to store document data in-memory rather than on disk? This way the performance can be increased because of reduced IO.

score 4 · Answer 1 · answered Jun 18 '20 at 10:37

4

You can either use attribute fields in combination with fast-search (to build posting lists on top of the data), or consider putting the index on RAM-disks. The most common approach is to use a combination of attribute fields and on-disk index and use SSD drives, though.

Also, for document summaries it is good for performance to have all fields in the most commonly requested summary class be attribute fields.

answered Jun 18 '20 at 10:37

Yngve Aasheim

71
2

Note also that Vespa will automatically use free memory to cache disk data. – Jon Jun 20 '20 at 08:48

In-memory document storage in Vespa.ai

1 Answers1