Re:We've saved 5000% in cloud costs by writing our own database
by swasheck on Saturday April 06, 2024 @09:16PM
(#39957266)
wholly unimpressed with the aurora line but my mgmt has swallowed it hook, line, and sinker
Re:We've saved 5000% in cloud costs by writing our own database
by sgarland on Sunday April 07, 2024 @09:21AM
(#39960576)
Mine thinks we’re going to be able to shift a massive, resource-hungry PG BDR mesh onto Aurora. I don’t buy it, purely from a performance capability standpoint. The only reason the shit queries we have run at all is through sheer performance.
SANs add latency, who knew? What did surprise me during testing is that Aurora seems to have added latency even if the entire dataset could easily fit into buffers, with zero disk reads required.
If you want actually fast cloud DBaaS, the only way is with instances that have local NVMe for caching, like AWS ??gd.db instances.