A blow by Murphy

[Rob Tito]

hello

Due to a flaw in the design of a database - one field I never ever expected people to use to search in - I had to index that particular field (being the only NOT indexed field)

The table contains some 4 million recprds. Indexing that one field took me about 3 days, as in 72 hours.

YET About 1 hour short of finishing - yup Murphy struck: MAC OS X 10.4.6 totally froze. Corrupted the filesystem of one of my hard disks so that took another two days to get restored again.

Now my question:

You can either export the field or table to another database and then import it while setting indexing on input on.

That however will take somewhat like another three days, and even though I can duck Murophy hits when least expected.

Are there accelerators for indexing?

Or is there a safer way to do it, maybe partwise or such.

Well any serious increase in sopeed will be greatly appreciated.

TIA

Rob

[xochi]

Hi Rob,

Is this file hosted with FileMaker Server, or are you opening it single-user using FileMaker Client? There are some cases where FMS performs way better than FMC, and of course, many cases where the opposite is true.

One performance optimization I've stumbled across -- the performance when indexing a field will be abysmal if the field index itself is larger than your RAM cache. I think in this case FM may actually have to save the entire index to disc with every single record it indexes. If you can get the entire file / table into RAM, you'll see the best performance.

Also, if the field is text based, you should experiment with "Minimum" vs. "All" indexing (on the Storage Tab under Field Definitions)

If you have the time, try a quick 2x2 matrix of re-indexing tests:

FMC using small RAM cache

FMC using giant RAM cache

FMS using small RAM cache (on both server & client)

FMS using giant RAM cache (on both server & client).

I'm sure many folks would be interested to hear your results.