![]() | |
![]() |
| | Thread Tools | Display Modes |
#1
| |||
| |||
|
#2
| |||
| |||
|
|
Hi, we have migrate our cubes from our 32 bit olap 2000 machine to x64 bit 2005. But after migrating seems 2005 perform worse performance in many cases that 2000. On our new server we have 8CPU and 64 GB of memory, i gave 20 GB to sql server and leave all SSAS memory options default because MS does not recomend change memory settings. Our cube files takes about 35 GB, but database Estimated Size from cube properties shows only 206 MB. Cube is partitioned. Looking at perfom, I see that Total queries from file is about 2000, Total queries from cache filtered 500000, Total queries from cache direct 30000. So, only few percent of queries answered from files, all others - from cache. In task manager msmdsrv takes 2 GB of memory. But in perfom MSAS 2005: Cache shows only 700 MB of memory and is the same as Memory AggcacheKB. Is it OK? Also, in the same MSAS 2005: Misses/sec is about 50 percent of Lookups/sec. Is it normal? So, every second query pass cache? Why? Also, interesting case with counter MSAS 2005: Proc aggregations memory size bytes. I used this perfom in SQL 2000 to see is there any job creating aggregates, especially it was usefull when I did incremental dimensions update. In SQL 2005 this counter shows nonzero only when partitions process are in "aggregate" state. But when I do incremental update for dimensions this counter always shows 0. Is it normal behaviour? One more thing - do I need to specify "slice" property for the partition? I read abouty this option I need to set only for ROLAP partitions. I tried to set slice property, but the same result. I missed way to known how many partitions is partcipating into olap query. Please, is there any suggestions how to optimize SSAS 2005 cubes. Ramunas Balukonis |
#3
| |||
| |||
|
|
Hi, we have migrate our cubes from our 32 bit olap 2000 machine to x64 bit 2005. But after migrating seems 2005 perform worse performance in many cases that 2000. On our new server we have 8CPU and 64 GB of memory, i gave 20 GB to sql server and leave all SSAS memory options default because MS does not recomend change memory settings. Our cube files takes about 35 GB, but database Estimated Size from cube properties shows only 206 MB. Cube is partitioned. Looking at perfom, I see that Total queries from file is about 2000, Total queries from cache filtered 500000, Total queries from cache direct 30000. So, only few percent of queries answered from files, all others - from cache. In task manager msmdsrv takes 2 GB of memory. But in perfom MSAS 2005: Cache shows only 700 MB of memory and is the same as Memory AggcacheKB. Is it OK? Also, in the same MSAS 2005: Misses/sec is about 50 percent of Lookups/sec. Is it normal? So, every second query pass cache? Why? Also, interesting case with counter MSAS 2005: Proc aggregations memory size bytes. I used this perfom in SQL 2000 to see is there any job creating aggregates, especially it was usefull when I did incremental dimensions update. In SQL 2005 this counter shows nonzero only when partitions process are in "aggregate" state. But when I do incremental update for dimensions this counter always shows 0. Is it normal behaviour? One more thing - do I need to specify "slice" property for the partition? I read abouty this option I need to set only for ROLAP partitions. I tried to set slice property, but the same result. I missed way to known how many partitions is partcipating into olap query. Please, is there any suggestions how to optimize SSAS 2005 cubes. Ramunas Balukonis |
#4
| |||
| |||
|
|
insure that your dimensions are correctly optimized and you use user hierarchy. use the rigid relationship, and keep the flexible option only as the last option. set the aggregationusage to full for your most requested attributes. In my case this step really improve the performance by setting up a better aggregation design. The slicer parameter is recommended but optionnal for your partitions can you describe in which cases the performance is bad? how many time the query takes? (in AS2000 & 2005) do you use calculated members? do you have distinct count measures? if yes, does these DCount are stored in a dedicated measuregroup? (1 by dcount measure) "Ramunas Balukonis" <ramblk2 (AT) hotmail (DOT) com> wrote in message news:1163056194.603024 (AT) loger (DOT) vpmarket.int... Hi, we have migrate our cubes from our 32 bit olap 2000 machine to x64 bit 2005. But after migrating seems 2005 perform worse performance in many cases that 2000. On our new server we have 8CPU and 64 GB of memory, i gave 20 GB to sql server and leave all SSAS memory options default because MS does not recomend change memory settings. Our cube files takes about 35 GB, but database Estimated Size from cube properties shows only 206 MB. Cube is partitioned. Looking at perfom, I see that Total queries from file is about 2000, Total queries from cache filtered 500000, Total queries from cache direct 30000. So, only few percent of queries answered from files, all others - from cache. In task manager msmdsrv takes 2 GB of memory. But in perfom MSAS 2005: Cache shows only 700 MB of memory and is the same as Memory AggcacheKB. Is it OK? Also, in the same MSAS 2005: Misses/sec is about 50 percent of Lookups/sec. Is it normal? So, every second query pass cache? Why? Also, interesting case with counter MSAS 2005: Proc aggregations memory size bytes. I used this perfom in SQL 2000 to see is there any job creating aggregates, especially it was usefull when I did incremental dimensions update. In SQL 2005 this counter shows nonzero only when partitions process are in "aggregate" state. But when I do incremental update for dimensions this counter always shows 0. Is it normal behaviour? One more thing - do I need to specify "slice" property for the partition? I read abouty this option I need to set only for ROLAP partitions. I tried to set slice property, but the same result. I missed way to known how many partitions is partcipating into olap query. Please, is there any suggestions how to optimize SSAS 2005 cubes. Ramunas Balukonis |
#5
| |||
| |||
|
|
Jerome, I'll try recomendations about rigid and especally AggregationUssage (I forgot about this). flexible/rigid are alomst the same as changing dimension type on AS 2000? I use lots of calc members, and some of them performs worse performance in AS 2005, especially with SUM function. Today I installed SQL server SP2 CTP on our DEV machine and performance dramatically grew up! Also, warning about unnatural hierarchies is great in SP2 CTP! But who known, when SP2 will be realeased? This is clue to our problems. Ramunas Balukonis "Jeje" <willgart (AT) hotmail (DOT) com> wrote in message news:EDD87352-4AEC-46C4-B99D-70A3EE5B79D6 (AT) microsoft (DOT) com... insure that your dimensions are correctly optimized and you use user hierarchy. use the rigid relationship, and keep the flexible option only as the last option. set the aggregationusage to full for your most requested attributes. In my case this step really improve the performance by setting up a better aggregation design. The slicer parameter is recommended but optionnal for your partitions can you describe in which cases the performance is bad? how many time the query takes? (in AS2000 & 2005) do you use calculated members? do you have distinct count measures? if yes, does these DCount are stored in a dedicated measuregroup? (1 by dcount measure) "Ramunas Balukonis" <ramblk2 (AT) hotmail (DOT) com> wrote in message news:1163056194.603024 (AT) loger (DOT) vpmarket.int... Hi, we have migrate our cubes from our 32 bit olap 2000 machine to x64 bit 2005. But after migrating seems 2005 perform worse performance in many cases that 2000. On our new server we have 8CPU and 64 GB of memory, i gave 20 GB to sql server and leave all SSAS memory options default because MS does not recomend change memory settings. Our cube files takes about 35 GB, but database Estimated Size from cube properties shows only 206 MB. Cube is partitioned. Looking at perfom, I see that Total queries from file is about 2000, Total queries from cache filtered 500000, Total queries from cache direct 30000. So, only few percent of queries answered from files, all others - from cache. In task manager msmdsrv takes 2 GB of memory. But in perfom MSAS 2005: Cache shows only 700 MB of memory and is the same as Memory AggcacheKB. Is it OK? Also, in the same MSAS 2005: Misses/sec is about 50 percent of Lookups/sec. Is it normal? So, every second query pass cache? Why? Also, interesting case with counter MSAS 2005: Proc aggregations memory size bytes. I used this perfom in SQL 2000 to see is there any job creating aggregates, especially it was usefull when I did incremental dimensions update. In SQL 2005 this counter shows nonzero only when partitions process are in "aggregate" state. But when I do incremental update for dimensions this counter always shows 0. Is it normal behaviour? One more thing - do I need to specify "slice" property for the partition? I read abouty this option I need to set only for ROLAP partitions. I tried to set slice property, but the same result. I missed way to known how many partitions is partcipating into olap query. Please, is there any suggestions how to optimize SSAS 2005 cubes. Ramunas Balukonis |
#6
| |||
| |||
|
|
SP2 will arrive shortly, its a prerequisite for Office 2007 and office 2007 is here!!! "Ramunas Balukonis" <ramblk2 (AT) hotmail (DOT) com> wrote in message news:1163428161.975861 (AT) loger (DOT) vpmarket.int... Jerome, I'll try recomendations about rigid and especally AggregationUssage (I forgot about this). flexible/rigid are alomst the same as changing dimension type on AS 2000? I use lots of calc members, and some of them performs worse performance in AS 2005, especially with SUM function. Today I installed SQL server SP2 CTP on our DEV machine and performance dramatically grew up! Also, warning about unnatural hierarchies is great in SP2 CTP! But who known, when SP2 will be realeased? This is clue to our problems. Ramunas Balukonis "Jeje" <willgart (AT) hotmail (DOT) com> wrote in message news:EDD87352-4AEC-46C4-B99D-70A3EE5B79D6 (AT) microsoft (DOT) com... insure that your dimensions are correctly optimized and you use user hierarchy. use the rigid relationship, and keep the flexible option only as the last option. set the aggregationusage to full for your most requested attributes. In my case this step really improve the performance by setting up a better aggregation design. The slicer parameter is recommended but optionnal for your partitions can you describe in which cases the performance is bad? how many time the query takes? (in AS2000 & 2005) do you use calculated members? do you have distinct count measures? if yes, does these DCount are stored in a dedicated measuregroup? (1 by dcount measure) "Ramunas Balukonis" <ramblk2 (AT) hotmail (DOT) com> wrote in message news:1163056194.603024 (AT) loger (DOT) vpmarket.int... Hi, we have migrate our cubes from our 32 bit olap 2000 machine to x64 bit 2005. But after migrating seems 2005 perform worse performance in many cases that 2000. On our new server we have 8CPU and 64 GB of memory, i gave 20 GB to sql server and leave all SSAS memory options default because MS does not recomend change memory settings. Our cube files takes about 35 GB, but database Estimated Size from cube properties shows only 206 MB. Cube is partitioned. Looking at perfom, I see that Total queries from file is about 2000, Total queries from cache filtered 500000, Total queries from cache direct 30000. So, only few percent of queries answered from files, all others - from cache. In task manager msmdsrv takes 2 GB of memory. But in perfom MSAS 2005: Cache shows only 700 MB of memory and is the same as Memory AggcacheKB. Is it OK? Also, in the same MSAS 2005: Misses/sec is about 50 percent of Lookups/sec. Is it normal? So, every second query pass cache? Why? Also, interesting case with counter MSAS 2005: Proc aggregations memory size bytes. I used this perfom in SQL 2000 to see is there any job creating aggregates, especially it was usefull when I did incremental dimensions update. In SQL 2005 this counter shows nonzero only when partitions process are in "aggregate" state. But when I do incremental update for dimensions this counter always shows 0. Is it normal behaviour? One more thing - do I need to specify "slice" property for the partition? I read abouty this option I need to set only for ROLAP partitions. I tried to set slice property, but the same result. I missed way to known how many partitions is partcipating into olap query. Please, is there any suggestions how to optimize SSAS 2005 cubes. Ramunas Balukonis |
#7
| |||
| |||
|
|
Correction: SP2 is not precisely a pre-requisite for Office 2007 -- but without SP2 some of the new Excel 2007 features against AS will be disabled. SP2 should become available around the first quarter of 2007. Thanks, Akshai -- Try out the MSDN Forums for Analysis Services at: http://forums.microsoft.com/MSDN/Sho...ID=83&SiteID=1 This posting is provided "AS IS" with no warranties, and confers no rights Please do not send email directly to this alias. This alias is for newsgroup purposes only. "Jeje" <willgart (AT) hotmail (DOT) com> wrote in message news 990EC88-C25C-4E0F-B613-89D53993DD41 (AT) microsoft (DOT) com...SP2 will arrive shortly, its a prerequisite for Office 2007 and office 2007 is here!!! "Ramunas Balukonis" <ramblk2 (AT) hotmail (DOT) com> wrote in message news:1163428161.975861 (AT) loger (DOT) vpmarket.int... Jerome, I'll try recomendations about rigid and especally AggregationUssage (I forgot about this). flexible/rigid are alomst the same as changing dimension type on AS 2000? I use lots of calc members, and some of them performs worse performance in AS 2005, especially with SUM function. Today I installed SQL server SP2 CTP on our DEV machine and performance dramatically grew up! Also, warning about unnatural hierarchies is great in SP2 CTP! But who known, when SP2 will be realeased? This is clue to our problems. Ramunas Balukonis "Jeje" <willgart (AT) hotmail (DOT) com> wrote in message news:EDD87352-4AEC-46C4-B99D-70A3EE5B79D6 (AT) microsoft (DOT) com... insure that your dimensions are correctly optimized and you use user hierarchy. use the rigid relationship, and keep the flexible option only as the last option. set the aggregationusage to full for your most requested attributes. In my case this step really improve the performance by setting up a better aggregation design. The slicer parameter is recommended but optionnal for your partitions can you describe in which cases the performance is bad? how many time the query takes? (in AS2000 & 2005) do you use calculated members? do you have distinct count measures? if yes, does these DCount are stored in a dedicated measuregroup? (1 by dcount measure) "Ramunas Balukonis" <ramblk2 (AT) hotmail (DOT) com> wrote in message news:1163056194.603024 (AT) loger (DOT) vpmarket.int... Hi, we have migrate our cubes from our 32 bit olap 2000 machine to x64 bit 2005. But after migrating seems 2005 perform worse performance in many cases that 2000. On our new server we have 8CPU and 64 GB of memory, i gave 20 GB to sql server and leave all SSAS memory options default because MS does not recomend change memory settings. Our cube files takes about 35 GB, but database Estimated Size from cube properties shows only 206 MB. Cube is partitioned. Looking at perfom, I see that Total queries from file is about 2000, Total queries from cache filtered 500000, Total queries from cache direct 30000. So, only few percent of queries answered from files, all others - from cache. In task manager msmdsrv takes 2 GB of memory. But in perfom MSAS 2005: Cache shows only 700 MB of memory and is the same as Memory AggcacheKB. Is it OK? Also, in the same MSAS 2005: Misses/sec is about 50 percent of Lookups/sec. Is it normal? So, every second query pass cache? Why? Also, interesting case with counter MSAS 2005: Proc aggregations memory size bytes. I used this perfom in SQL 2000 to see is there any job creating aggregates, especially it was usefull when I did incremental dimensions update. In SQL 2005 this counter shows nonzero only when partitions process are in "aggregate" state. But when I do incremental update for dimensions this counter always shows 0. Is it normal behaviour? One more thing - do I need to specify "slice" property for the partition? I read abouty this option I need to set only for ROLAP partitions. I tried to set slice property, but the same result. I missed way to known how many partitions is partcipating into olap query. Please, is there any suggestions how to optimize SSAS 2005 cubes. Ramunas Balukonis |
![]() |
| Thread Tools | |
| Display Modes | |
| |