Making use of multiple cores?

[databaser]

I think I recently read somewhere that Filemaker can only use one core on a machine and gets no benefit from multiple cores.

I'm constantly running long-running scripts that perform intense calculations on large amounts of data, and I was interested in investing in a custom-built 32-48 core windows server machine, hoping it would dramatically reduce the time it takes these scripts to run (some take 18 to 90 seconds; I was hoping that could get down to 0.1 to 2 seconds. Some take around 24 hours; I was hoping to get those down to 15-120 minutes).

So does anyone know, how many cores (if even plural) can filemaker take advantage of? Aside from the obvious idea of trying to improve the scripts (something I've spent many hours doing, and there may be little more room for improvement) are there other ways to reduce script-running time (I was mostly thinking better hardware, but am interested in any recommendations)?

Thanks in advance.

[David Jondreau]

I don't know anything about how Filemaker uses cores, but I do know script steps run in serial and I think you want a script to run in parallel. These scripts are being performed locally, right?

I would focus on the scripting. There may be big time savers you're missing. Loading your data into and doing calc in variables, for example, is generally a lot faster than fields.

[IdealData]

long-running scripts that perform intense calculations on large amounts of data

Isn't that what FM calculation fields are for?

AFAIK FileMaker SERVER is multi-threaded, so if you were using that and all the script steps were supported then you might gain some advantage - but I still think your DBs are not yet fully designed if you are running scripts in this way.

[Vaughan]

Use the uber-box to run multiple vms, each running a robot FMP client.

[IdealData]

Uber-box Vaughan?

What is uber-box?. I'm all into VMs but can't find this one. I know about Parallels (server) and VirtualBox.

Nice "thinking outside the box" approach to get the unused cores into action though.

[Vaughan]

Uber-box is any big server computer. (Uber=German) :D

[databaser]

Thanks for the virtual machine idea.

On using calculation fields, I'm pretty sure that would just make everything much slower.

I did write filemaker about this. They ask that you not publish their responses, so I won't copy and paste it, but they essentially said that 4 cores can be used for performing calculation, and that they haven't tested this on more than 4 cores (sounds strange that they wouldn't have tested that more thoroughly, but that's what the response said)

[Vaughan]

On using calculation fields, I'm pretty sure that would just make everything much slower.

So perform a test and find out, rather than just guess.

[databaser]

I think know from experience,

when you're doing lots of calculations that are very taxing, you need to be in complete control of when the the calculations are being performed, otherwise the database is unmanageable.

I think this is especially so when you're using complex algorithms to parse text (especially when the algorithms have to very complex to accomodate minimal uniformity in the text).