Hello, Community.
I'm wondering if ruleloadfromfile has significant affect on data load performance. If we load data into cube with heavy feeders (long chain of dependent cubes) i see data load speed is awfully bad, but if I put ruleloadfromfile(dummy file) into prolog and epilog (original rule file) I get normal speed of data load.
Could you test is it really true please?
Or maybe I'm confused by something else.
Last edited by ioscat on Tue Apr 16, 2013 11:35 am, edited 1 time in total.
ioscat wrote:Hello, Community.
I'm wondering if ruleloadfromfile has significant affect on data load performance. If we load data into cube with heavy feeders (long chain of dendent cubes) i see data load speed is awfully bad, but if I put ruleloadfromfile(dummy file) into prolog and epilog (original rule file) I get normal speed of data load.
Could you test is it really true please?
Or maybe I'm confused by something else.
It is not so much RuleLoadFromFile but the rules themselves and more specifically the feeders.
The thing that affects data loading throughput is the number of feeders being fired when new data hits a numeric leaf cell. If you effectively turn the feeders (and rules) off by loading a blank rule file before the load then obviously the data loading throughput will be much better, exactly how much better will depend on the degree of feeding and maybe overfeeding and whether there is any additional complexity of feeding a rule calculated cell that then feeds another and then another and so on versus just a straight feed. It is relatively common that you can achieve better over all cycle time for big data loads by first removing rules, then loading data, then reapplying rules. Whether it is worth all the extra effort will depend on the time saving up for grabs and the processing window that is available.
so i'm interested is the time saving is extremely significant or it is just incidental, is it widely used by other TM1 developers or maybe it is called micro-optimization or worst-practice.
well, in our case we use the same trick for second time. but we still worry about ability to loose rules during process.
