Are time-based index faster than rollover-indices - if the time is known?

[G-K-Patel]

Hi All,

We know there are two types mechanism to create multiple indices, if the size of data is too large for one index. Namely, 1) time-based: where time is used in the suffix-of the index name, e.g. MyIndex-2019, MyIndex-2020, etc. & 2) rollover-indices: which is the new approach as discussed here

My question is, which is faster given we know the Time-Range in which the user wants to search ? For example, if I know the user wants to search data in 2019, in time-based index I would direct the query to MyIndex-2019, but in rollover-index I am forced to query all indices even if I know the time.

Ofcourse it is understandable, that the first query is slow for the rollover-index, but will the sebsequent queries also be slow ? Do you have any benchmark comparing performace of time-based and rollover indices.

Thanks for your time and support.

[Christian_Dahlqvist]

Have you measured the difference? Querying indices that do not have any relevant data should be quite efficient. If you have tried, did the indices queried have the same amount of data?

[G-K-Patel]

Not really. But, I thought you guys might have some test-bench results. I needed this information so that I can properly engineer my Indexing strategy. My only problem is to know, if the search would get slower when the number of indices being queried increases over time. In case of time based index, I have to possibility to direct query to specific Indices (based on user selected time) but in case of rolling index I dont have that possibility.