Re: Raw I/O vs files
Date: Wed, 28 Jul 1993 15:16:55 GMT
Message-ID: <1993Jul28.151655.9306_at_zed.com>
In article <2500_at_coyote.UUCP>, gainer_at_slowmo.almaden.ibm.com (P. Gainer) writes:
|>
|> Has anyone done much benchmarking to determine whether recent raw I/O
|> implementations are *significantly* faster than file system implementations?
About a year ago I did some benchmarking with Informix-ONLINE 5.0 on an HP9000/730 to determine if it was worth using raw I/O. To my amazement, cooked files outperformed raw I/O by about 10%. At this point I called Informix tech support, and was informed that on some hardware platforms, the UNIX file system could outperform raw I/O, and that HP was one such platform.
In a development environment, I would not hesitate to use cooked files. However, if mission critical data is involved, I would always choose raw I/O. Using cooked files can result in data inconsistency after recovery. See the Online Administrator's Guide for a detailed explanation. (page 1-38)
-- -------------------------------------------------------------------------- Curt Leaist |Internet: curtl_at_zed.com Database Administrator |UUCP : ...!uunet!uupsi2!aruba!curtl --------------------------------------------------------------------------Received on Wed Jul 28 1993 - 17:16:55 CEST