To mirror or not to mirror

mirroringmemory
2 posts / 0 new
Last post
Tarjei Kirkesaether A+ 58
TK
To mirror or not to mirror

In terms of system usage, how much extra processing power does keeping modules mirrored add?

I am in the process of updating my debt section and have created separate modules for each debt and hedging instrument type (fixed, floating bank debt etc.). I am keeping the debt modules and swap modules mirrored while I undertake the testing back to my old numbers, however the model has increased 50% in size (30MB to 45MB) and it is now running a A LOT slower than it used to. I have also added 180 (15 years) columns to my time series to undertake all calculations on a monthly basis (cash interest is a real pain when moving from monthly to annual within the same time series!) so I'm worried this is the real culprit of the reduced performance but before I go and rebuild my modules in a shorter time series I would just like to know if I can expect the un-mirroring of the modules to significantly improve performance?

PS: Before anyone suggest the 'linked workbook' approach this is not an option for me in this instance:)

Michael Hutchens A+ 189

Hi Tarjei - It looks like you're continuing to play the role of Excel memory limits test pilot!

Re mirroring modules - there is no performance or memory different between inserting a module, duplicating a module or mirroring a module, but for the fact that when changes are made to non-assumptions within a mirrored module, they are automatically aligned across all mirrored modules. This can take time, but usually saves a bunch of time, so it's technically a net performance improvement, if you get what I mean!

Re 180 time series columns - my guess is that this is almost certainly the driver behind your performance issues. Assuming you're using at least some lookup-type functions (such as SUMIF, SUMPRODUCT, INDEX, OFFSET, etc.), this large number of columns will be making Excel do hundreds of thousands, if not millions, of looping calculations every time you change something. So it's going to blow out your file size, and your real-time performance, considerably.

Also, if you're talking 45MB in *.xlsb format, that's a MASSIVE Excel file, as it equates to over 100MB in *.xlsx (i.e. non-binary) format. Excel file format affects only size on disk though, not performance once loaded, it's just something I've raised because I'd always aimed to keep things below 20MB *.xlsx!

Something I always do when I'm building large models is try releasing them from Modano and then using them released. This will make it clear whether Modano is adding real-time overhead, which it really shouldn't be doing in 99% of cases. And if it is, we're always keen to do some performance testing on the scenario.