"Don't mean sheeit!"
One of many
And that could be true with Amazon's latest silicon.
The AWS Graviton4 processor packs 96 cores that offer on average 30% higher compute performance compared to Graviton3 and is 40% faster in database applications as well as 45% faster in Java applicationsIt kinda, sorta depends on what they mean by 'database applications', now doesn't it? If they mean, and I strongly suspect they do, just faster RBAR reads, then that's just shit. Coders have been out performing RDBMS engines from the beginning, continue to do so (COBOL/COPYBOOK), and with good reason; they can 'monitor' their transactions however poorly they wish. Codd's notion was that data and it's integrity were not separable, really. At the time, COBOL and VSAM were lingua franca in (mainframe, 99.99%) commercial applications. IBM helped along with CICS as transaction processing monitor, so coders didn't have to write their own transaction engine (almost always as an assembler module; yuck) if they didn't really, really have to.
But still, today, esp. among the java/web kiddies, RDBMS (MySQL, SS, et al), it's all about 'doing the transactions in the client'. Stupid is as stupid does. With innterTubes bandwidth being what it is, server bandwidth being what it is, server memory capacity being what it is, there's no justificable reason for being that stupid. Admit that the 'client' (mostly some PC) is nothing more than a GUI-fied VT100, and keep all that logic with the data where it belongs. Reduce the cpu load everywhere, and (horrors!) reduce the code load everywhere. But it does keep lots more java/web coders writing lots more code that would be oodles more efficient in 3/5 NF schemas. Oh well.
No comments:
Post a Comment