![]() | |
![]() |
| | Thread Tools | Display Modes |
#1
| |||
| |||
|
#2
| |||
| |||
|
|
HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#3
| |||
| |||
|
|
a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#4
| |||
| |||
|
|
a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#5
| |||
| |||
|
|
It's sad to hear that the MDX queries on a single partition cannot be paralelised ![]() The disk sys on the xeon srv is a FC SAN where I have an array of 4 FC HDD in RAID 10... quite fast. What is bothering me are the XMLA commands. For example, saving aggregations (XMLA) on the PC is almost twice as fast as on the XEON server ... Xeon 2.8GHz, PC 3.2 GHz... so in this case GHz > Mutiprocesor ![]() "Jéjé" <willgart_A_ (AT) hotmail_A_ (DOT) com> wrote in message news:eSz1rT$qGHA.4508 (AT) TK2MSFTNGP04 (DOT) phx.gbl... a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#6
| |||
| |||
|
|
It is not accurate that a single partition query cannot be parallelized. It is the formula engine that is serialized in evaluating calculations. Some metadata operations will also be serialized (like create/alter/delete). However, your description below is a little confusing to me -- do you find that it is only the metadata operations (create/alter/process) that are slow on the x64 box. Are queries fast? That may indicate network issues if just processing is slow... Are the number of databases on the x64 box equivalent to the 32-bit box? And the environments? Perhaps Profiler would help narrow things down a little more. One thing you could try is to install a new x86 instance on the x64 box and see how slow/fast it works -- if it is also slow, then that would indicate an environment/hardware issue. Also, keep in mind that 64-bit is not expected to necessarily give you better performance -- but rather better scalability. In fact, an x64 application has to deal with double pointer sizes and this can cause some perf degradation in some cases. In other cases, the larger physical memory available makes better caching possible which helps performance of queries. 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:eLuV0b$qGHA.4336 (AT) TK2MSFTNGP02 (DOT) phx.gbl... It's sad to hear that the MDX queries on a single partition cannot be paralelised ![]() The disk sys on the xeon srv is a FC SAN where I have an array of 4 FC HDD in RAID 10... quite fast. What is bothering me are the XMLA commands. For example, saving aggregations (XMLA) on the PC is almost twice as fast as on the XEON server ... Xeon 2.8GHz, PC 3.2 GHz... so in this case GHz > Mutiprocesor ![]() "Jéjé" <willgart_A_ (AT) hotmail_A_ (DOT) com> wrote in message news:eSz1rT$qGHA.4508 (AT) TK2MSFTNGP04 (DOT) phx.gbl... a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#7
| |||
| |||
|
|
Hi, The query performance is quite good. My issues are related with XMLA command execution (metadata operations) like create/modify aggregation design, roles a.s.o. The server is dual dual-core with HT .(8 virt exec units). When an XMLA is executed one of the execution units goes 100% for some seconds (10-20). For example I've just redesigned the aggregations for a MOLAP partition using the wizard. At the last step (pressing Finish) the window freeze for (10-20) sec and the msmdsrv.exe process going 100% on one execution unit. I've used the management studio on the server via remote desktop. I've taked the backup of the OLAP database, restore-it to a PC (3.2Ghz P4, x86) and make the same operation wich there takes 1-2 sec. All XMLA commands are somehow very slow executed on the server. The SLQ is SP1 build 2153. The Win2003 OS is up to date with SP1 and all online updates available. Thanks, Radu "Akshai Mirchandani [MS]" <akshaim (AT) online (DOT) microsoft.com> wrote in message news:OlxE6e1rGHA.516 (AT) TK2MSFTNGP05 (DOT) phx.gbl... It is not accurate that a single partition query cannot be parallelized. It is the formula engine that is serialized in evaluating calculations. Some metadata operations will also be serialized (like create/alter/delete). However, your description below is a little confusing to me -- do you find that it is only the metadata operations (create/alter/process) that are slow on the x64 box. Are queries fast? That may indicate network issues if just processing is slow... Are the number of databases on the x64 box equivalent to the 32-bit box? And the environments? Perhaps Profiler would help narrow things down a little more. One thing you could try is to install a new x86 instance on the x64 box and see how slow/fast it works -- if it is also slow, then that would indicate an environment/hardware issue. Also, keep in mind that 64-bit is not expected to necessarily give you better performance -- but rather better scalability. In fact, an x64 application has to deal with double pointer sizes and this can cause some perf degradation in some cases. In other cases, the larger physical memory available makes better caching possible which helps performance of queries. 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:eLuV0b$qGHA.4336 (AT) TK2MSFTNGP02 (DOT) phx.gbl... It's sad to hear that the MDX queries on a single partition cannot be paralelised ![]() The disk sys on the xeon srv is a FC SAN where I have an array of 4 FC HDD in RAID 10... quite fast. What is bothering me are the XMLA commands. For example, saving aggregations (XMLA) on the PC is almost twice as fast as on the XEON server ... Xeon 2.8GHz, PC 3.2 GHz... so in this case GHz > Mutiprocesor ![]() "Jéjé" <willgart_A_ (AT) hotmail_A_ (DOT) com> wrote in message news:eSz1rT$qGHA.4508 (AT) TK2MSFTNGP04 (DOT) phx.gbl... a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#8
| |||
| |||
|
|
It sounds like the difference is unrelated to query performance and disk i/o etc. My guess is that it is something environmental -- perhaps you have lots of active sessions on the 64-bit server which causes transaction commit to be an expensive operation because it has to refresh active sessions. Or the cached data on the 64-bit server is much larger and it the commit operation is releasing memory pages... Or perhaps you have lots of other databases on the 64-bit server that are not present on the 32-bit server. Have you tried to restart the AS process on the 64-bit box? Also, I would suggest taking a look at the event log to see if there are unusual errors/warnings reported there... Another option to try would be to rename your current data folder and start from a new data folder, restore the backup file and work from there... 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:OCWuSv7rGHA.4580 (AT) TK2MSFTNGP05 (DOT) phx.gbl... Hi, The query performance is quite good. My issues are related with XMLA command execution (metadata operations) like create/modify aggregation design, roles a.s.o. The server is dual dual-core with HT .(8 virt exec units). When an XMLA is executed one of the execution units goes 100% for some seconds (10-20). For example I've just redesigned the aggregations for a MOLAP partition using the wizard. At the last step (pressing Finish) the window freeze for (10-20) sec and the msmdsrv.exe process going 100% on one execution unit. I've used the management studio on the server via remote desktop. I've taked the backup of the OLAP database, restore-it to a PC (3.2Ghz P4, x86) and make the same operation wich there takes 1-2 sec. All XMLA commands are somehow very slow executed on the server. The SLQ is SP1 build 2153. The Win2003 OS is up to date with SP1 and all online updates available. Thanks, Radu "Akshai Mirchandani [MS]" <akshaim (AT) online (DOT) microsoft.com> wrote in message news:OlxE6e1rGHA.516 (AT) TK2MSFTNGP05 (DOT) phx.gbl... It is not accurate that a single partition query cannot be parallelized. It is the formula engine that is serialized in evaluating calculations. Some metadata operations will also be serialized (like create/alter/delete). However, your description below is a little confusing to me -- do you find that it is only the metadata operations (create/alter/process) that are slow on the x64 box. Are queries fast? That may indicate network issues if just processing is slow... Are the number of databases on the x64 box equivalent to the 32-bit box? And the environments? Perhaps Profiler would help narrow things down a little more. One thing you could try is to install a new x86 instance on the x64 box and see how slow/fast it works -- if it is also slow, then that would indicate an environment/hardware issue. Also, keep in mind that 64-bit is not expected to necessarily give you better performance -- but rather better scalability. In fact, an x64 application has to deal with double pointer sizes and this can cause some perf degradation in some cases. In other cases, the larger physical memory available makes better caching possible which helps performance of queries. 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:eLuV0b$qGHA.4336 (AT) TK2MSFTNGP02 (DOT) phx.gbl... It's sad to hear that the MDX queries on a single partition cannot be paralelised ![]() The disk sys on the xeon srv is a FC SAN where I have an array of 4 FC HDD in RAID 10... quite fast. What is bothering me are the XMLA commands. For example, saving aggregations (XMLA) on the PC is almost twice as fast as on the XEON server ... Xeon 2.8GHz, PC 3.2 GHz... so in this case GHz Mutiprocesor ![]() "Jéjé" <willgart_A_ (AT) hotmail_A_ (DOT) com> wrote in message news:eSz1rT$qGHA.4508 (AT) TK2MSFTNGP04 (DOT) phx.gbl... a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#9
| |||
| |||
|
|
It sounds like the difference is unrelated to query performance and disk i/o etc. My guess is that it is something environmental -- perhaps you have lots of active sessions on the 64-bit server which causes transaction commit to be an expensive operation because it has to refresh active sessions. Or the cached data on the 64-bit server is much larger and it the commit operation is releasing memory pages... Or perhaps you have lots of other databases on the 64-bit server that are not present on the 32-bit server. Have you tried to restart the AS process on the 64-bit box? Also, I would suggest taking a look at the event log to see if there are unusual errors/warnings reported there... Another option to try would be to rename your current data folder and start from a new data folder, restore the backup file and work from there... 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:OCWuSv7rGHA.4580 (AT) TK2MSFTNGP05 (DOT) phx.gbl... Hi, The query performance is quite good. My issues are related with XMLA command execution (metadata operations) like create/modify aggregation design, roles a.s.o. The server is dual dual-core with HT .(8 virt exec units). When an XMLA is executed one of the execution units goes 100% for some seconds (10-20). For example I've just redesigned the aggregations for a MOLAP partition using the wizard. At the last step (pressing Finish) the window freeze for (10-20) sec and the msmdsrv.exe process going 100% on one execution unit. I've used the management studio on the server via remote desktop. I've taked the backup of the OLAP database, restore-it to a PC (3.2Ghz P4, x86) and make the same operation wich there takes 1-2 sec. All XMLA commands are somehow very slow executed on the server. The SLQ is SP1 build 2153. The Win2003 OS is up to date with SP1 and all online updates available. Thanks, Radu "Akshai Mirchandani [MS]" <akshaim (AT) online (DOT) microsoft.com> wrote in message news:OlxE6e1rGHA.516 (AT) TK2MSFTNGP05 (DOT) phx.gbl... It is not accurate that a single partition query cannot be parallelized. It is the formula engine that is serialized in evaluating calculations. Some metadata operations will also be serialized (like create/alter/delete). However, your description below is a little confusing to me -- do you find that it is only the metadata operations (create/alter/process) that are slow on the x64 box. Are queries fast? That may indicate network issues if just processing is slow... Are the number of databases on the x64 box equivalent to the 32-bit box? And the environments? Perhaps Profiler would help narrow things down a little more. One thing you could try is to install a new x86 instance on the x64 box and see how slow/fast it works -- if it is also slow, then that would indicate an environment/hardware issue. Also, keep in mind that 64-bit is not expected to necessarily give you better performance -- but rather better scalability. In fact, an x64 application has to deal with double pointer sizes and this can cause some perf degradation in some cases. In other cases, the larger physical memory available makes better caching possible which helps performance of queries. 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:eLuV0b$qGHA.4336 (AT) TK2MSFTNGP02 (DOT) phx.gbl... It's sad to hear that the MDX queries on a single partition cannot be paralelised ![]() The disk sys on the xeon srv is a FC SAN where I have an array of 4 FC HDD in RAID 10... quite fast. What is bothering me are the XMLA commands. For example, saving aggregations (XMLA) on the PC is almost twice as fast as on the XEON server ... Xeon 2.8GHz, PC 3.2 GHz... so in this case GHz Mutiprocesor ![]() "Jéjé" <willgart_A_ (AT) hotmail_A_ (DOT) com> wrote in message news:eSz1rT$qGHA.4508 (AT) TK2MSFTNGP04 (DOT) phx.gbl... a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
#10
| |||
| |||
|
|
I've traced with profiler the XMLA sent to AS. I've modified a role diving it Admin priviledges. The management studio sent 2 XMLA to as wich total execution time was around 15 sec. I was alone on the server. Each execution made one execution unit 100%. In the meantime the server was not able to respond queries sent to other database. As soon as the XMLA was executed the queries get answered. I have AS 2005 EE x64., 2x processor license. Thanks, Radu --- first Alter AllowCreate="true" ObjectExpansion="ObjectProperties" xmlns="http://schemas.microsoft.com/analysisservices/2003/engine" Object DatabaseID>ROEFIX_PROD</DatabaseID RoleID>Einkauf</RoleID /Object ObjectDefinition Role xmlns:xsd="http://www.w3.org/2001/XMLSchema" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" ID>Einkauf</ID Name>Einkauf</Name Members Member Name>RRO\escharin</Name /Member Member Name>RRO\agut</Name /Member /Members /Role /ObjectDefinition /Alter PropertyList xmlns="urn:schemas-microsoft-com:xml-analysis" Timeout>0</Timeout SspropInitAppName>Microsoft SQL Server Management Studio</SspropInitAppName LocaleIdentifier>3079</LocaleIdentifier ClientProcessID>2076</ClientProcessID /PropertyList --- second Alter AllowCreate="true" ObjectExpansion="ObjectProperties" xmlns="http://schemas.microsoft.com/analysisservices/2003/engine" Object DatabaseID>ROEFIX_PROD</DatabaseID DatabasePermissionID>Einkauf</DatabasePermissionID /Object ObjectDefinition DatabasePermission xmlns:xsd="http://www.w3.org/2001/XMLSchema" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" ID>Einkauf</ID Name>Einkauf</Name RoleID>Einkauf</RoleID Process>true</Process ReadDefinition>Allowed</ReadDefinition Read>Allowed</Read Administer>true</Administer /DatabasePermission /ObjectDefinition /Alter PropertyList xmlns="urn:schemas-microsoft-com:xml-analysis" Timeout>0</Timeout SspropInitAppName>Microsoft SQL Server Management Studio</SspropInitAppName LocaleIdentifier>3079</LocaleIdentifier ClientProcessID>2076</ClientProcessID /PropertyList "Akshai Mirchandani [MS]" <akshaim (AT) online (DOT) microsoft.com> wrote in message news:eU7qMyAsGHA.1272 (AT) TK2MSFTNGP05 (DOT) phx.gbl... It sounds like the difference is unrelated to query performance and disk i/o etc. My guess is that it is something environmental -- perhaps you have lots of active sessions on the 64-bit server which causes transaction commit to be an expensive operation because it has to refresh active sessions. Or the cached data on the 64-bit server is much larger and it the commit operation is releasing memory pages... Or perhaps you have lots of other databases on the 64-bit server that are not present on the 32-bit server. Have you tried to restart the AS process on the 64-bit box? Also, I would suggest taking a look at the event log to see if there are unusual errors/warnings reported there... Another option to try would be to rename your current data folder and start from a new data folder, restore the backup file and work from there... 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:OCWuSv7rGHA.4580 (AT) TK2MSFTNGP05 (DOT) phx.gbl... Hi, The query performance is quite good. My issues are related with XMLA command execution (metadata operations) like create/modify aggregation design, roles a.s.o. The server is dual dual-core with HT .(8 virt exec units). When an XMLA is executed one of the execution units goes 100% for some seconds (10-20). For example I've just redesigned the aggregations for a MOLAP partition using the wizard. At the last step (pressing Finish) the window freeze for (10-20) sec and the msmdsrv.exe process going 100% on one execution unit. I've used the management studio on the server via remote desktop. I've taked the backup of the OLAP database, restore-it to a PC (3.2Ghz P4, x86) and make the same operation wich there takes 1-2 sec. All XMLA commands are somehow very slow executed on the server. The SLQ is SP1 build 2153. The Win2003 OS is up to date with SP1 and all online updates available. Thanks, Radu "Akshai Mirchandani [MS]" <akshaim (AT) online (DOT) microsoft.com> wrote in message news:OlxE6e1rGHA.516 (AT) TK2MSFTNGP05 (DOT) phx.gbl... It is not accurate that a single partition query cannot be parallelized. It is the formula engine that is serialized in evaluating calculations. Some metadata operations will also be serialized (like create/alter/delete). However, your description below is a little confusing to me -- do you find that it is only the metadata operations (create/alter/process) that are slow on the x64 box. Are queries fast? That may indicate network issues if just processing is slow... Are the number of databases on the x64 box equivalent to the 32-bit box? And the environments? Perhaps Profiler would help narrow things down a little more. One thing you could try is to install a new x86 instance on the x64 box and see how slow/fast it works -- if it is also slow, then that would indicate an environment/hardware issue. Also, keep in mind that 64-bit is not expected to necessarily give you better performance -- but rather better scalability. In fact, an x64 application has to deal with double pointer sizes and this can cause some perf degradation in some cases. In other cases, the larger physical memory available makes better caching possible which helps performance of queries. 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. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:eLuV0b$qGHA.4336 (AT) TK2MSFTNGP02 (DOT) phx.gbl... It's sad to hear that the MDX queries on a single partition cannot be paralelised ![]() The disk sys on the xeon srv is a FC SAN where I have an array of 4 FC HDD in RAID 10... quite fast. What is bothering me are the XMLA commands. For example, saving aggregations (XMLA) on the PC is almost twice as fast as on the XEON server ... Xeon 2.8GHz, PC 3.2 GHz... so in this case GHz Mutiprocesor ![]() "Jéjé" <willgart_A_ (AT) hotmail_A_ (DOT) com> wrote in message news:eSz1rT$qGHA.4508 (AT) TK2MSFTNGP04 (DOT) phx.gbl... a standard MDX query against a non partitionned cube use only 1 thread so 1 CPU. if you have partitions or a query which ask for measures from different measure group you should see more then 1 CPU to works at the same time. its true, sometimes a single PC is better then a Xeon server, but not when you start to have multiple users or processes. sorry to say this, but I also found SATA drives far better then SCSI drives!!! SCSI drives are good only when the controller has enough cache. so take a look at your disks and test them. "Radu Colceriu" <radu_colceriu (AT) hotmail (DOT) com> wrote in message news:%23IxPW18qGHA.1852 (AT) TK2MSFTNGP03 (DOT) phx.gbl... HI.. I have an setup like - 2xXeon 2.8 dual-core 8GRAM, Win2003 x64, SQL 2005 EE x64 - PC P4 3.2G 2 G RAM. Win 2003 x86 SQL 2005 EE x86 I don't know why but on the PC the AS is working *much* faster than on the Xeon system. Especially designig agregation/creating roles/ creating partitions a.s.o. So it seems that the same XMLA is executed almost twice as fast on the PC than on the dual xeon. On the dual xeon maschine each time an XMLA is executed, one of the 8 (2xdualcore + hyperthreading) "cores" is 100% but all other are doing nothing. On the same maschines SQL is running much faster on the Xeon maschine than on the PC. Any ideeas? Thanks, Radu |
![]() |
| Thread Tools | |
| Display Modes | |
| |