Big block transfers: good or bad?

Posted Apr 1, 2004 3:01 UTC (Thu) by sbergman27 (guest, #10767)
Parent article: Big block transfers: good or bad?

So with the obliteration of the 128k ata write limit, will we finally be able to turn off hardware write caching without a performance hit. Turning off write caching is necessary if you really want the safety benefits promised by jounalling file systems. Even ext3 in data=journal mode is not truly safe with write caching left on. With the current state of affairs, we either put up with danger or turn off write caching and accept a big write performance penalty.

Big block transfers: good or bad?

Posted Apr 1, 2004 4:23 UTC (Thu) by dlang (guest, #313) [Link]

no, this won't solve the write-cache problem. there was another thread in the last week or so about write barriers for IDE that is working on that issue