Hi Spiral,
If I were you, I would absolutely consider using table partitioning. There
are a couple of questions to be answered.
1. What is the rate/speed of the table's growth?
2. What is the range of values you use for mid columns to query the table?
Are they generally close to each other? Or,
On Sun, Jan 08, 2023 at 07:02:01AM -0500, spiral wrote:
> This table is used essentially as a key-value store; rows are accessed
> only with `mid` primary key. Additionally, inserted rows may only be
> deleted, but never updated.
>
> We only run the following queries:
> - INSERT INTO messages
That’s crazy only having 8GB memory when you have tables with over 100GBs. One
general rule of thumb is have enough memory to hold the biggest index.
Sent from my iPad
> On Jan 9, 2023, at 3:23 AM, spiral wrote:
>
> Hello,
>
> We have a table containing ~1.75 billion rows, using 170GB
Hello,
We have a table containing ~1.75 billion rows, using 170GB storage.
The table schema is the following:
messages=# \d messages
Table "public.messages"
Column| Type | Collation | Nullable | Default
--+-+---+--+-
mid