Database Internals - SlateDB with Chris Riccomini

[@thrawn01]

Thank you for this, looking forward to assisting with the golang port! This was a great overview of the project, and a great way to get acquainted with some of the design decisions.

[@TheGeekNarrator]

Thanks for watching 🙏🏻

[@ShortGiant1]

Great video, learnt a lot!

[@Jasin-p6w]

Wonderful video

[@nosh3019]

Thanks for this interview. Great as always! One question: so if the WAL writes are batched and periodically flushed to storage, there in a window in which if the process crashes, the writes will be lost? I.e the client response that write is complete is not durable.

[@TheGeekNarrator]

Thanks for watching. Yes, there are ways to configure the behaviour. This might help you slatedb.io/docs/faq/#what-happens-if-the-process-goes-down-before-slatedb-flushes-data-to-object-storage

[@ShortGiant1]

Yeah that’s the “linger” stuff they spoke about. The client won’t get an ack about the write being successful until the flush happens. Therefore it would be as if the write request never happened (in practice the client would retry)

[@jmitesh01]

Information dense conversation! And isn't it expected that size is sufficiently large then it would increase number of put calls to object storage. Why don't you do buffering based on time plus size based criteria?

[@TheGeekNarrator]

Good question. I think there are plans but haven’t really finalised yet. But maybe in the future a hybrid setting will be available. IIRC I asked this question to Chris.

[@anthonya880]

Please remove members only videos. It doesn't make sense yet for your channel.

[@TheGeekNarrator]

Thanks for the feedback. I am happy you added “yet”, which tells you believe 😀🙏🏻

[@anthonya880]

@@TheGeekNarrator Yes your channel is different. It has lot of potential.