Обсуждение: efficiency random values / sequential ID values in indexes
Hi
Is there any appreciable difference in using random values or sequential values in indexes
in a multi tenanted application there is a choice that the single field ID's value is totally random / UUID or the numbers are created with a prefix of the tenant. Since all access will be within a tenant only, will it make any performance difference between using purely random values vs <tenant no prefix part>+<random value>.
thanks
Sanjay
On Mon, Apr 15, 2024 at 6:05 AM Sanjay Minni <sanjay.minni@gmail.com> wrote:
HiIs there any appreciable difference in using random values or sequential values in indexesin a multi tenanted application there is a choice that the single field ID's value is totally random / UUID or the numbers are created with a prefix of the tenant. Since all access will be within a tenant only, will it make any performance difference between using purely random values vs <tenant no prefix part>+<random value>.
Two benefits of <tenant no prefix part>+<random value>:
1. In a non-partitioned table, it gives your index "locality of data": all of customer X's record pointers are in This subtree. Makes buffers more efficient when a customer runs reports. Bonus points if you then regularly CLUSTER using that table.
2. Makes table partitioning by <tenant prefix> much easier. That also enhances locality of data.
Just make sure that the field ID is BIGINT...
Thanks
On Mon, Apr 15, 2024 at 6:19 PM Ron Johnson <ronljohnsonjr@gmail.com> wrote:
On Mon, Apr 15, 2024 at 6:05 AM Sanjay Minni <sanjay.minni@gmail.com> wrote:HiIs there any appreciable difference in using random values or sequential values in indexesin a multi tenanted application there is a choice that the single field ID's value is totally random / UUID or the numbers are created with a prefix of the tenant. Since all access will be within a tenant only, will it make any performance difference between using purely random values vs <tenant no prefix part>+<random value>.Two benefits of <tenant no prefix part>+<random value>:1. In a non-partitioned table, it gives your index "locality of data": all of customer X's record pointers are in This subtree. Makes buffers more efficient when a customer runs reports. Bonus points if you then regularly CLUSTER using that table.2. Makes table partitioning by <tenant prefix> much easier. That also enhances locality of data.Just make sure that the field ID is BIGINT...