|
|
You're in the
sql server data warehouse
group:Varchar vs Nvarchar
sql server data warehouse:
I have a database that contains a table with 33.6 million records. When I built the table I used Nvarchar in several of the fields. Now I'm thinking that because the db size is 7.5 gb, that I could reduce that by changing the datatype to Varchar. Since I'm not changing languages or porting hardware platforms is there a good reason why I should keep the nvarchar data type other that it will take an eternity to save the new table structure? Does anyone have any suggestions as to which would be faster, building a new table and importing the records via a query, or just changing the existing datatype and saving the table. TIA! Mark
I guess I've always felt smaller & tighter dbs are more efficient than larger
dbs. It's also a slow and timely process to transfer the backup to my laptop where I do some development on. Thanks for the comments. Mark [quoted text, click to view] "Adam Machanic" wrote:
> While I agree that you might not need NVARCHAR based on your description, > I'm not sure that it's something worth worrying about. 7.5 gigs is not a > lot of disk space... Why the concern? > > > -- > Adam Machanic > Pro SQL Server 2005, available now > http://www.apress.com/book/bookDisplay.html?bID=457 > -- > > > "MChrist" <MChrist@discussions.microsoft.com> wrote in message > news:5653976D-FA2D-4E0C-AECA-DDEF7640E124@microsoft.com... > > Please pardon my ignorance here, but I'm a newbie with MS SQL Server. > > > > I have a database that contains a table with 33.6 million records. When I > > built the table I used Nvarchar in several of the fields. Now I'm > > thinking > > that because the db size is 7.5 gb, that I could reduce that by changing > > the > > datatype to Varchar. > > > > Since I'm not changing languages or porting hardware platforms is there a > > good reason why I should keep the nvarchar data type other that it will > > take > > an eternity to save the new table structure? > > > > Does anyone have any suggestions as to which would be faster, building a > > new > > table and importing the records via a query, or just changing the existing > > datatype and saving the table. > > > > TIA! > > > > Mark > > > > > >
While I agree that you might not need NVARCHAR based on your description,
I'm not sure that it's something worth worrying about. 7.5 gigs is not a lot of disk space... Why the concern? -- Adam Machanic Pro SQL Server 2005, available now http://www.apress.com/book/bookDisplay.html?bID=457 -- [quoted text, click to view] "MChrist" <MChrist@discussions.microsoft.com> wrote in message news:5653976D-FA2D-4E0C-AECA-DDEF7640E124@microsoft.com... > Please pardon my ignorance here, but I'm a newbie with MS SQL Server. > > I have a database that contains a table with 33.6 million records. When I > built the table I used Nvarchar in several of the fields. Now I'm > thinking > that because the db size is 7.5 gb, that I could reduce that by changing > the > datatype to Varchar. > > Since I'm not changing languages or porting hardware platforms is there a > good reason why I should keep the nvarchar data type other that it will > take > an eternity to save the new table structure? > > Does anyone have any suggestions as to which would be faster, building a > new > table and importing the records via a query, or just changing the existing > datatype and saving the table. > > TIA! > > Mark > >
[quoted text, click to view] "MChrist" <MChrist@discussions.microsoft.com> wrote in message news:66C53E73-90D7-4196-AB6F-1FD09C880E33@microsoft.com... >I guess I've always felt smaller & tighter dbs are more efficient than >larger > dbs. It's also a slow and timely process to transfer the backup to my > laptop > where I do some development on. > I have to agree here. Although 7.6g is really not much, a database half that size is significantly easier to manage and will perform better automatically. David
"David Browne" <davidbaxterbrowne no potted meat@hotmail.com> wrote in
message news:ODq7FGE9FHA.444@TK2MSFTNGP11.phx.gbl... [quoted text, click to view] > > I have to agree here. Although 7.6g is really not much, a database half > that size is significantly easier to manage and will perform better > automatically. True. Smaller data means fewer data pages to read into memory when retrieving rows, which means fewer I/O operations, which means better performance. But the flip side is how long it might take to convert the entire database... and if there's no performance problem already, it may not be worthwhile. -- Adam Machanic Pro SQL Server 2005, available now http://www.apress.com/book/bookDisplay.html?bID=457 --
To get back to the original question. If he decides to change datatype, what
is faster? Altering the table or creating new/importing data? My guess would be that altering the table would be a better solution. MC [quoted text, click to view] "Adam Machanic" <amachanic@hotmail._removetoemail_.com> wrote in message news:OW60XmE9FHA.2264@tk2msftngp13.phx.gbl... > "David Browne" <davidbaxterbrowne no potted meat@hotmail.com> wrote in > message news:ODq7FGE9FHA.444@TK2MSFTNGP11.phx.gbl... >> >> I have to agree here. Although 7.6g is really not much, a database half >> that size is significantly easier to manage and will perform better >> automatically. > > True. Smaller data means fewer data pages to read into memory when > retrieving rows, which means fewer I/O operations, which means better > performance. But the flip side is how long it might take to convert the > entire database... and if there's no performance problem already, it may > not be worthwhile. > > > -- > Adam Machanic > Pro SQL Server 2005, available now > http://www.apress.com/book/bookDisplay.html?bID=457 > -- > > >
[quoted text, click to view]
"MC" <marko_culo#@#yahoo#.#com#> wrote in message news:OVfaiQL9FHA.3636@TK2MSFTNGP09.phx.gbl... > To get back to the original question. If he decides to change datatype, > what is faster? Altering the table or creating new/importing data? > My guess would be that altering the table would be a better solution. Well, the answer according to this -very simple- test is that creating the new table is somewhat faster (but this test doesn't take things like indexes into account -- so YMMV): --- use tempdb go SET NOCOUNT ON create table x (blah nvarchar(400)) go insert x (blah) select top 5000 replicate('x', 400) from master..spt_values a, master..spt_values b go declare @starttime datetime set @starttime = getdate() alter table x alter column blah varchar(400) PRINT datediff(ms, @starttime, getdate()) GO DROP TABLE x GO create table x (blah nvarchar(400)) go insert x (blah) select top 5000 replicate('x', 400) from master..spt_values a, master..spt_values b go declare @starttime datetime set @starttime = getdate() create table y (blah nvarchar(400)) insert y (blah) select blah from x drop table x exec sp_rename 'y', 'x' PRINT datediff(ms, @starttime, getdate()) GO drop table x go --- -- Adam Machanic Pro SQL Server 2005, available now http://www.apress.com/book/bookDisplay.html?bID=457 --
Yeah, I'd like to know myself. I'm fairly new to SQL Server too and just
had a huge database thrown at me to manage. Some queries are now timing out so performance is key. It's worth it to me to convert fields where it will help. Randall Arnold [quoted text, click to view] "MC" <marko_culo#@#yahoo#.#com#> wrote in message news:OVfaiQL9FHA.3636@TK2MSFTNGP09.phx.gbl... > To get back to the original question. If he decides to change datatype, > what is faster? Altering the table or creating new/importing data? > My guess would be that altering the table would be a better solution. > > MC > > > "Adam Machanic" <amachanic@hotmail._removetoemail_.com> wrote in message > news:OW60XmE9FHA.2264@tk2msftngp13.phx.gbl... >> "David Browne" <davidbaxterbrowne no potted meat@hotmail.com> wrote in >> message news:ODq7FGE9FHA.444@TK2MSFTNGP11.phx.gbl... >>> >>> I have to agree here. Although 7.6g is really not much, a database half >>> that size is significantly easier to manage and will perform better >>> automatically. >> >> True. Smaller data means fewer data pages to read into memory when >> retrieving rows, which means fewer I/O operations, which means better >> performance. But the flip side is how long it might take to convert the >> entire database... and if there's no performance problem already, it may >> not be worthwhile. >> >> >> -- >> Adam Machanic >> Pro SQL Server 2005, available now >> http://www.apress.com/book/bookDisplay.html?bID=457 >> -- >> >> >> > >
have you try to partition your table horizontally?
create 1 table with the most requested columns, and a second table with your "slow" varchar columns. The row size in the first table will be smaller, so you'll have more rows by page. but, because you ask under the data warehouse newsgroup, remember that reducing the number of joins improve the response time. [quoted text, click to view] "Adam Machanic" <amachanic@hotmail._removetoemail_.com> wrote in message news:ehtCA5T9FHA.132@TK2MSFTNGP15.phx.gbl... > "MC" <marko_culo#@#yahoo#.#com#> wrote in message > news:OVfaiQL9FHA.3636@TK2MSFTNGP09.phx.gbl... >> To get back to the original question. If he decides to change datatype, >> what is faster? Altering the table or creating new/importing data? >> My guess would be that altering the table would be a better solution. > > > > Well, the answer according to this -very simple- test is that creating > the new table is somewhat faster (but this test doesn't take things like > indexes into account -- so YMMV): > > --- > use tempdb > go > > SET NOCOUNT ON > > create table x (blah nvarchar(400)) > go > > insert x (blah) > select top 5000 > replicate('x', 400) > from > master..spt_values a, > master..spt_values b > go > > declare @starttime datetime > set @starttime = getdate() > > alter table x > alter column blah varchar(400) > > PRINT datediff(ms, @starttime, getdate()) > GO > > DROP TABLE x > GO > > create table x (blah nvarchar(400)) > go > > insert x (blah) > select top 5000 > replicate('x', 400) > from > master..spt_values a, > master..spt_values b > go > > declare @starttime datetime > set @starttime = getdate() > > create table y (blah nvarchar(400)) > > insert y (blah) > select blah > from x > > drop table x > > exec sp_rename 'y', 'x' > > PRINT datediff(ms, @starttime, getdate()) > GO > > drop table x > go > --- > > > -- > Adam Machanic > Pro SQL Server 2005, available now > http://www.apress.com/book/bookDisplay.html?bID=457 > -- > >
[quoted text, click to view] "Jéjé" <willgart@BBBhotmailAAA.com> wrote in message news:OOvqXIN%23FHA.2472@TK2MSFTNGP12.phx.gbl... > have you try to partition your table horizontally? > create 1 table with the most requested columns, and a second table with > your "slow" varchar columns. > > The row size in the first table will be smaller, so you'll have more rows > by page. > > but, because you ask under the data warehouse newsgroup, remember that > reducing the number of joins improve the response time. > This is almost always a bad idea, and in SQL Server 2005 (with indexes with included columns) the "almost" goes away. You can do better by adding commonly used columns to the clustered index (still not always a good idea, but better than "horizontal partitioning". David
[quoted text, click to view] "Adam Machanic" <amachanic@hotmail._removetoemail_.com> wrote in message news:u8D1jeR%23FHA.2708@TK2MSFTNGP12.phx.gbl... > "David Browne" <davidbaxterbrowne no potted meat@hotmail.com> wrote in > message news:%232u2stO%23FHA.1988@TK2MSFTNGP12.phx.gbl... >> >> This is almost always a bad idea, and in SQL Server 2005 (with indexes >> with included columns) the "almost" goes away. > > Don't forget row overflow! This still doesn't let us design tables > with an unlimited number of columns, but the number goes way up. > > >> You can do better by adding commonly used columns to the clustered index >> (still not always a good idea, but better than "horizontal partitioning". > > I think you mean "vertical"? Horizontal, per my understanding, is the > kind supported by partitioned views in SQL Server 2000 and table/index > partitioning in SQL Server 2005 -- and that kind of partitioning is a good > thing, IMO. > > Yes vertical. I was lying down when I wrote that. David
"David Browne" <davidbaxterbrowne no potted meat@hotmail.com> wrote in
message news:%232u2stO%23FHA.1988@TK2MSFTNGP12.phx.gbl... [quoted text, click to view] > > This is almost always a bad idea, and in SQL Server 2005 (with indexes > with included columns) the "almost" goes away. Don't forget row overflow! This still doesn't let us design tables with an unlimited number of columns, but the number goes way up. [quoted text, click to view] > You can do better by adding commonly used columns to the clustered index > (still not always a good idea, but better than "horizontal partitioning". I think you mean "vertical"? Horizontal, per my understanding, is the kind supported by partitioned views in SQL Server 2000 and table/index partitioning in SQL Server 2005 -- and that kind of partitioning is a good thing, IMO. -- Adam Machanic Pro SQL Server 2005, available now http://www.apress.com/book/bookDisplay.html?bID=457 --
Don't see what you're looking for? Try a search.
|
January 2003
March 2003
July 2003
August 2003
September 2003
October 2003
November 2003
December 2003
January 2004
February 2004
March 2004
April 2004
May 2004
June 2004
July 2004
August 2004
September 2004
October 2004
November 2004
December 2004
January 2005
February 2005
March 2005
April 2005
May 2005
June 2005
July 2005
August 2005
September 2005
October 2005
November 2005
December 2005
January 2006
February 2006
March 2006
April 2006
May 2006
June 2006
July 2006
August 2006
September 2006
October 2006
November 2006
December 2006
January 2007
February 2007
March 2007
April 2007
May 2007
June 2007
July 2007
August 2007
September 2007
October 2007
November 2007
| home · blog · groups | Questions? Comments? Contact the dn |