About 3 years ago, my coworker "Charles" convinced 14 people in my org to join a Fantasy Football league with a $100 buy-in. (14 people pay $100 at the beginning, and then the $1,400 would be distributed to the top 3 teams). I was not one of these people. It was a bit much for an office pool, ...
I think there are two questions here:
Why the relational model?
Why apply normalization principles when designing a relational database?
A1. In answer to the first question, the relational model is an extremely flexible and powerful abstraction of data. It supports data integrity rules, data ...
I left a transaction open for 24 hours, did a whole lot of log-generating stuff, restarted the instance, and it recovered immediately, with all data fully available for querying/updating. Quite like magic really.
just a couple of weeks ago we had a vendor-provided upgrade utility that timed-out after 30 minutes, necessitating around 40 minutes of rollback every time it timed out. Lovely.
One of the great things about the implementation is how orthogonal it is. Things like AGs, distributed transactions and whatnot just work.
@swasheck Yep
So long as indirect checkpoints are set up reasonably frequently, failover/recovery ought to be much faster.
There's a little extra log generated (but no more flushes) so it's reasonably close to a free lunch. I'd much rather pay an ongoing small overhead than wait 18 hours for rollback/recovery anyway. Won't suit every workload, but then nothing does.
i'm still a bit fuzzy on managed instance and how that breaks down differently from a DBA tweaking and tuning perspective, but that's because i've invested nearly 0 time into reading about it