Re: Q: Mirroring - what is the better way

In article <19971112051101.AAA17902@ladder02.news.aol.com>, VFraenkle > <vfraenkle@aol.com> writes > >I have heard many different statements about mirroring. The one say > >"You allways have to use OS mirroring" - the other say "It is better > > to belive in INFORMIX mirroring". > > > >An here is my final question, now: > > > >INFORMIX build in mirroring vs OS mirroring, what is the better way > > and why? > > > >Thank you, Volker David Williams responded: > According to the person at Informix UK Training who gave my > Performance Tuning course....there is no difference. Mirroring is > mirroring. > > YOu have to remember some OS's do not including OS mirroring. On > these machines Informix mirroring is the only choice. > > I suspect Informix only developed their mirroring because OS's > at the time did not include mirroring... Allow me to insert my $.02 US: I have always figured on OS mirroring of the raw devices to be faster because it would be handled at the device driver level, even with logical volumes. However, this comes with a certain inflexibility because the OS expects the primary volume and the mirror volume to be identical, sector by sector. With OnLine mirroring, I can mirror one dbspace (like the one containing my logs), skip mirroring another dbspace (like my temp space) although both might be on the same raw [logical] volume. Flexibility is useful at times. As for David's remark about some OS's not including mirroring, i'd like to qualify this statement: I recall some machines that would do mirroring only on file systems, never on raw devices. (I forget which these were.) Thus, the only option becomes OnLine mirroring. -- -- Jake (Never yelled "CROWDED THEATER!" during a fire) +------------------------------------------------------------+ | The expedient performance of a task with excessive concern | | regarding its duration-to-completion engenders a virtual | | certainty of diminished benefit therefrom. | | -- Benjamin Franklin (but he said it in 3 words) | +------------------------------------------------------------+