Geek asks and answers

Section Strategy

Starting point :
I fill one table with data, and to compare data from run to run, I added the "run" column (Integer).
I keep the last 5 starts. Since I continue to get timeouts after 2 runs in the table, I decided to split the table into "run".

What I did :
1. I created a partitioned table with 1000 partitions (I assume that "run" will be from 1 to 1000), there are a maximum of 5 partitions, the remaining 995 are empty and wait until the next run is launched.
2. I created a partition scheme that uses the partition function described above and saves all partitions (that is, 1000) in the PRIMARY file group.
3. I split the table into "run".

Result: there is no longer a timeout, but the processing time has increased by 50%.

Question : Which points were suboptimal in my partitioning strategy? Should I create 5 partitions and change the way I calculate mileage?

+3
source share
1 answer

Separation is not a solution to performance problems. You are looking for indexes . Partitioning will slow down processing, and best of all, you can hope this is table-level performance. Partitioning is a good fit for data management scenarios, for quickly moving data using a partition switch (ETL or data expiration).

, -. (.. ). .

+6

More articles:


All Articles