Hacker News new | past | comments | ask | show | jobs | submit login

I find it hard to believe this actually caused a DB problem, updating an int column on an indexed table should have taken nanoseconds. Far faster than any HTTP calls should have been happening. Sounds like something was very wrong. Someone fucked up with the network causing some severe DB latency or they'd written some really shitty code.

Even if it did, as the other comment says, you could easily work round it. If that was the behaviour there was no good reason to have it utterly perfect.




Depends if autocommit is enabled. If it is (and on many DBs it's the default) then you're telling the DB that by the time the increment completes, it should be written to durable storage. That's a minimum of one seek, which on a hard disk is about 4ms, which = max 250 TPS. On a SSD you might get 0.1ms seek times = 10k TPS.

It's possible to kill your performance very quickly with databases if you request features that you don't actually need.


That's where decisions around consistency kick in.

You need to decide if you absolutely do not want to serve any API call unless you are sure they've been paid for, in which case you have to create a commit transaction on that account for each call. Or you decide how much you can let a would be rotten customer get away with, and uses queues which leads to eventual consistencies.




Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: