Problem with setting alias

[speedy]

Hello,

I have a dimension with 100.000 elements and want to set the alias for those elements using the API function TM1ObjectAttributeValueSet. With TM1 9.0 that was no problem. As we are upgrading to version 9.5, it takes about 10 times longer than before. I tried to use TM1ServerBatchUpdateStart and TM1EnableBulkLoadMode but that was no improvement in performance.

Code: Select all

voEle = TM1ObjectListHandleByIndexGet( pLocal, voDim, TM1DimensionElements(), TM1ValIndex( pLocal, 1 ) ) ;

voAttribute = TM1ObjectListHandleByNameGet( pLocal, voEle, TM1ObjectAttributes(), TM1ValString( pLocal, ALIAS_BEZEICHNUNG.c_str(), 0 ) ) ;

for ( i=0; i<TM1Element.Length; i++ )
{
    hEle = TM1ObjectListHandleByNameGet( pEle, voDim, TM1DimensionElements(), TM1ValString( pLocal, TM1Element[i].c_str(), 0 ) ); 
    TM1ObjectAttributeValueSet( pAttribute, hEle, voAttribute, TM1ValString( pLocal,  TM1Alias[i].c_str(), 0 ) ) ;
}

What can I do to get the same performance as before in version 9.0?

[Mike Cowie]

Hi,

First, if you're updating attributes like this, why do you need the TM1 API? Could you try a Turbo Integrator process, which has all the attribute-updating functionality you've listed here?

Second, if you've upgraded to a 9.5 version, is it possible your hardware specs and or network location have changed? It would be good to at least rule out a hit in performance because you're now taking a network route with more hops or latency.

Third, could you try updating the }ElementAttributes_ cube directly using the cube value update API functions?

Fourth, are you still using a 9.0 API version to connect to the 9.5 server? I'm not sure that would be a problem, but thought I'd ask.

If you rule out all the above, you should reach out to IBM with your performance statistics, particularly if you can hand them a snippet of API code and dimension with attributes that they can use to run in both 9.0 and 9.5 environments.

And... I'd still look to possible TI processes instead of API code since TI will run entirely on the TM1 server, eliminating the network layer inherent in API programming. If you must use the API then could you at least execute such a TI process using the TM1 API?

I noticed a definite hit in performance with view value retrieval in the TM1 API after 9.0, and that has gotten somewhat better since 9.1, but it has still not gotten back to 9.0 levels of performance. The TM1 server engine has changed significantly since 9.0 and I don't think you'll see the same performance in, say, 9.5 that you saw in 9.0 for operations like this, but the payoff is that performance should generally be much better with many concurrent users compared with 9.0 and prior. That being said, I wouldn't expect performance to be an order of magnitude slower, which is why you may want to report to IBM on this one, especially if you can give them something that demonstrates the problem.

Regards,
Mike