The often repeated recommendation to not exceed 8GB/25% of RAM for shared buffers in postgres is wrong. For pgbench scale 1500, on laptop with fast SSD and 32GB of RAM, parallelism of 16. s_b of 1GB, 8GB, 16GB, using huge pages. r/o: 115k, 100k, 185k r/w: 16300, 15100, 21500
-
Show this thread
-
Replying to @AndresFreundTec
Thanks for the results. A few questions: 1) what OS? If Linux: 2) is any of this a result of kernel VM improvements? 3) for r/w is there more response time variance with 16gb because of writeback storms?
1 reply 0 retweets 0 likes -
Replying to @MarkCallaghanDB
It's indeed linux. And I don't think it's really VM improvement depandant - I think this has been true for a long time. The response time for 24GB (see copy/paste correction upthread) is vastly better than for the other shared buffer settings, as there is no buffer replacement.
2 replies 0 retweets 1 like -
Replying to @AndresFreundTec @MarkCallaghanDB
I was recently informed on Slack that it was changes around 9.6 (or possibly 10) that reduced the old wisdom to hearsay.
1 reply 0 retweets 0 likes
I think it was largely wrong before then too - there were a few more caveats though.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.