Showing posts with label partition. Show all posts
Showing posts with label partition. Show all posts

Friday, February 24, 2012

ABOUT sql server 2005 partition

hello,

We are writting an erp for group,and want to use partition of sql server 2005 for large data,our environment is window 2003 server,8g memory.my one solution will create many physical files(about 500-600) to fit 500-600 partitions,how about performance? or I will use another arithmetic in order to reduce partitions?

Hi,

Wondering why you will need that many partitioons for. How big your table is likely to be ? how many rows? What are planning to use as you partition key?

Jag

|||

hello,

Thank you for your attention,following are our case:

we wrote an erp for a group,it have five child companys,three factorys,six trademark,we finish distribution sale,supply chain,production(not running),our database size now is 19G on sql server 2000,the system is slow,we now upgrading hardware and software,database will use sql server 2005.

I analyse the system, if I use partition to divide trademark,customer data and warehouse data and supplier data,producing,and every child company also will use partition,I think it will be better more,if a company have 10 partitons, five companys will reach 50 partitions,I really not know how many partitions will right,and performace is best?

I also want to do an online system for small company,if using design above,50 companys will up to 500 partitions,I think this case,the arithmetic will different from above,it will not grow with company amount,is right?

Thank you,again.

|||

It really seems like that is a lot of partitions, especially for a database that is only 19gb. Really, the idea behind using partitions is somewhat two-fold. First, you can segreate your archive data from your live data for performance reasons. Also, you can segment your historical or archive data itself for performance benefit. You can also move data from partition to partition as needed. However, there is a limit the number of partitions you can create, and as the number of partitions grows, the amount of administration grows. This leads to the type of hardware you are using.....are you using a SAN or some sort of high-performance shared disk storage? If so, what RAID level are you using? If this design is optimal, it might be time to look at how your application is written. Most likely, you will see the highest performance gain by slight design changes and query optimizations as compared to hardware upgrades and partitioning.

Tim

|||

hello,

thank you for your suggest,I learn a lot.

The company have two system,one for sale,one for purchase,we now will incorporate these system,incorporation size are at least 30g,retail system are working recently,all shops are about 400 amount,so data will grow faster than before,at next quater,the production system will begin,also will produce so much data,we use raid5,donn't use SAN,Perhaps I need study more knowledge about SAN.

we upgrade hardware and software,the old system is slow,often happen concurrentcy problem,the erp first is bought ,but couldn't fit the company,the architecture is very bad,we do a great lot of redevelopping,now I redesign architecture,consider optimization at transation,query,analyse,and in order to predigest the communication between company, customer, warehouse,factory,supplyer, I put some table together to one table,using partition to divide them,I think it will have good performance for transation, and persormance for group report,my design have consider about partition administrator, but as you have mentioned,partition for history data, yes,it will good for performance,there is another solution, I want to design an OLAP system for histotry analysing,is it right?

at last,how much about partition limit? thank you

Thursday, February 16, 2012

about partition?

I created partition for one table

How to check if the partitioned data is correct?

Also if there is easier and quick method to tune this partition?

any idea? many thanks

A first ideea is to use $partition function (in Books Online have an example); with it you can get the number of rows in each nonempty partition of a partitioned table or index; make a job that insert time to time this data into a table and read it periodically.

Gigi Ciubuc
www.sqlserver.ro|||

thank you

I am going to read BOL

about partition copy

hello,

I want to copy all or part data in one partition to another partition,how can I do?

do you mean a partitioned table partition? If so search out Kimberley Tripp's whitepaper on partitioning and read the sliding window scenario part. It's very easy at meta data level to move from one table to another.

|||

thank you for your answer, there have some difficulties,The database have hundred of partitioned tables,and have many relations,My purpose is to backup,export,import,copy some partition. How to do this efficienly,transationly?

|||

Your question is very general. Let me try to answer them here

(1) backup: lowest granularity is filegroup. There is no backup functionality at partition level. You can ofcourse create one dedicated filegroup for each partition to acchive it but it may not what you want to do

(2) You can always export using BCP all the rows in a specific partition.

(3) I believe BCP in also works on paritioned table.

|||thank you for your answer, (2)(3) will help me.

about partition copy

hello,

I want to copy all or part data in one partition to another partition,how can I do?

do you mean a partitioned table partition? If so search out Kimberley Tripp's whitepaper on partitioning and read the sliding window scenario part. It's very easy at meta data level to move from one table to another.

|||

thank you for your answer, there have some difficulties,The database have hundred of partitioned tables,and have many relations,My purpose is to backup,export,import,copy some partition. How to do this efficienly,transationly?

|||

Your question is very general. Let me try to answer them here

(1) backup: lowest granularity is filegroup. There is no backup functionality at partition level. You can ofcourse create one dedicated filegroup for each partition to acchive it but it may not what you want to do

(2) You can always export using BCP all the rows in a specific partition.

(3) I believe BCP in also works on paritioned table.

|||thank you for your answer, (2)(3) will help me.