Updatable Clustered ColumnStore Indexes in SQL Server 2014

Updatable Clustered ColumnStore Index that is introduced with SQL Server 2014. Before we go down to the details, I want to give you a brief overview about their first appearance in SQL Server 2012, and their limitations.

ColumnStore Indexes in SQL Server 2012:

The introduction of ColumnStore Indexes in SQL Server 2012 was one of the hottest new features (besides AlwaysOn). If chosen right, they can make an impressive performance improvement for Data Warehousing workloads. Unfortunately, they had 2 big limitations:

• There was only the support for Non-Clustered ColumnStore Indexes
• As soon as you had created a ColumnStore Index on a table, the underlying table was read only, and no changes to the data were allowed anymore
  

Both limitations were a huge bummer for some customers. Imagine you had a table with 300 GB of traditional row store data. With a ColumnStore Index it’s possible to compress that data down to a size of 30 GB. But SQL Server 2012 only allowed a Non-Clustered ColumnStore Index, means you had to store your data twice: one in the traditional row store format, and once in the new ColumnStore format. That’s a huge waste of storage, because your queries will (hopefully) only use your Non-Clustered ColumnStore Index.
And as soon as you had created your Non-Clustered ColumnStore Index, you were not allowed to change the underlying table data anymore – your table was just read only! Of course, there were some workarounds for this problem, like Partition Switching, but you still needed to implement that on your own…

ColumnStore Indexes in SQL Server 2014:

Things are changing now with SQL Server 2014, because Microsoft has resolved the above mentioned issues – with some magic and illusion: SQL Server 2014 provides you an Updatable Clustered ColumnStore Index! Let’s have a more detailed look on how this magic and illusion happens internally in SQL Server.
The first most important fact is that an underlying direct update of a ColumnStore Index is not possible! It would be too time consuming to do the complete decompress and compress on the fly during your INSERT, UPDATE, and DELETE transactions. Therefore SQL Server 2014 uses help from some magic: Delta Stores and Delete Bitmaps. Let’s have a more detailed look on both concepts.
Every time when you run an INSERT statement, the new record isn’t directly inserted into the ColumnStore Index – the record is inserted into a Delta Store. The Delta Store itself is nothing else than a traditional B-Tree structure with all its pro’s and con’s. When you afterwards read from the ColumnStore Index, SQL Server returns you the data from the compressed ColumnStore Index AND also from the Delta Store.
When you run an DELETE statement, again nothing happens in the compressed ColumnStore Index. The only thing that happens is that the record is deleted logically through a Delete Bitmap. Every record in the ColumnStore Index has a corresponding bit in that Delete Bitmap. When you again read your ColumnStore Index, SQL Server just discards the rows that are marked as deleted in the Delete Bitmap.
And running an UPDATE statement just means inserting the new version into the Delta Store, and marking the old version as deleted in the Delete Bitmap. Easy, isn’t it? The following picture (source http://research.microsoft.com/apps/pubs/default.aspx?id=193599) shows this concept in more details.

Because of the Delta Stores and the Delete Bitmap it seems that your ColumnStore Index is updatable, but in reality it is just immutable. There is also a background process called the Tuple Mover, which runs regularly and finally pushes your changes asynchronously into the compressed ColumnStore Index.

 In addition you can also finally define a ColumnStore Index in SQL Server 2014 as Clustered. This means that you don’t need to have your data in the traditional row store format anymore. Just create your table and create a Clustered ColumnStore Index on it. You can some with this approach huge space savings in your storage, because everything is now compressed. When you work in more details with the concepts of relational databases, Clustered always means Sorted. But with a Clustered ColumnStore Index this statement is not true: when you create a Clustered ColumnStore Index, there is NO sorting order in your data! Just be aware of this tiny little fact.

Buffer Pool Extensions in SQL Server 2014

Buffer Pool Extensions are introduced with SQL Server 2014. As you know, the Buffer Pool is one of the main memory consumers in SQL Server. When you read data from your storage, the data is cached in the Buffer Pool. SQL Server caches Execution Plans in the Plan Cache, which is also part of the Buffer Pool. The more physical memory you have, the larger your Buffer Pool will be (configured through the Max Server Memory setting).

A lot of SQL Server customers are dealing with the problem that physical memory is restricted in database servers: all memory slots are already occupied, so how do you want to add additional memory to the physical server? Of course, you can migrate to a larger server, but that’s another story… The solution to this specific problem is the introduction of Buffer Pool Extensions in SQL Server 2014. With the help of the Buffer Pool Extensions SQL Server introduces another layer in the memory hierarchy. Let’s have a look on the following picture:

As you can see you have on the top the Buffer Pool itself, which is very fast (regarding latency times), and at the bottom you can see our traditional storage, which is mainly slow. And the Buffer Pool Extensions are just settled down between of both – between the traditional Buffer Pool and our storage. The Buffer Pool Extensions itself are consisting of one simple file (the so-called Extension File) that should be stored on very fast storage – like a SSD drive. The Extension File is mainly the same as the page file in the Windows OS. Instead of adding additional physical memory to your database server, you just configure an Extension File on a SSD drive.

Before talking about the configuration and the concrete use of the Buffer Pool Extensions, I want to talk a little bit about the architecture and design behind the Buffer Pool Extensions. The traditional Buffer Pool of SQL Server always differentiates between clean and dirty pages. A clean page is a page which has the same content as the page in the storage. A dirty page is changed in memory, but hasn’t yet written to the storage. Around every minute the so-called CHECKPOINT process writes dirty pages out to the storage, means that our dirty page becomes a clean page.

The Buffer Pool Extensions itself are only used if the Buffer Pool of SQL Server get’s into memory pressure. Memory pressure means that SQL Server needs more memory than currently available. In that case the Buffer evicts pages from the Buffer Pool, which were least recently used. SQL Server uses here a Least Recently Used Policy (LRU). If you have now configured an Extension File, then SQL Server will write these pages into it, instead of writing them directly out to our slow storage. If the page is a dirty one, then the page will be also concurrently written to the physical storage (through an asynchronous I/O operation). Therefore you can’t loose any data when you are dealing with the Buffer Pool Extensions. At some point in time your Extension File will be also completely full. In that case SQL Server has to evict older pages from the Extension File (again through a LRU policy), and finally writes them to the physical storage. The Extension File just acts as an additional layer between the Buffer Pool and the storage itself.
Let’s have now a look on how we can configure the Buffer Pool Extensions in SQL Server 2014. SQL Server offers you here the command ALTER SERVER CONFIGURATION SET BUFFER POOL EXTENSION. Let’s have a more detailed look on how you use it:

ALTER SERVER CONFIGURATION
SET BUFFER POOL EXTENSION ON
(
   FILENAME = 'd:\ExtensionFile.BPE',
   SIZE = 15 GB
)
GO

The first restriction that you are hitting here is the fact that the Extension File must have at least the same size as the Buffer Pool itself. If you are specifying a smaller file size, you will get a lovely error message from SQL Server:

Msg 868, Level 16, State 1, Line 1
Buffer pool extension size must be larger than the current memory allocation threshold 1596 MB. Buffer pool extension is not enabled.

The next restriction that you will definitely hit is that you can’t change the size of the Extension File during the runtime of SQL Server. For example, when you want to change the Extension File to a larger size, you have to disable the Buffer Pool Extensions, and re-enable them again. You will have of course a performance degradation in the time of this operation, because you are just disabling one important caching layer used by SQL Server.

And in addition you are also not able to reduce the size of the Extension File, the file must be always larger than previously. Otherwise you are again getting an error message:

Msg 868, Level 16, State 1, Line 3
Buffer pool extension size must be larger than the current memory allocation threshold 4096 MB. Buffer pool extension is not enabled.

The whole configuration of the Buffer Pool Extensions can be also retrieved through the Dynamic Management View sys.dm_os_buffer_pool_extension_configuration.

Finding Last Backup Time for All Database – Last Full, Differential and Log Backup

SET NOCOUNT ON
GO
SET quoted_identifier OFF
DECLARE @dbname AS VARCHAR(80)
DECLARE @msgdb AS VARCHAR(100)
DECLARE @dbbkpname AS VARCHAR(80)
DECLARE @dypart1 AS VARCHAR(2)
DECLARE @dypart2 AS VARCHAR(3)
DECLARE @dypart3 AS VARCHAR(4)
DECLARE @currentdate AS VARCHAR(10)
DECLARE @server_name AS VARCHAR(30)
SELECT @server_name = @@servername
SELECT @dypart1 = DATEPART(dd,GETDATE())
SELECT @dypart2 = DATENAME(mm,GETDATE())
SELECT @dypart3 = DATEPART(yy,GETDATE())
SELECT @currentdate= @dypart1 + @dypart2 + @dypart3
PRINT "#####################################################################"
PRINT "# SERVERNAME : "+ @server_name + " DATE : "+ @currentdate +"#"
PRINT "#####################################################################"
PRINT "DatabaseName Full Diff TranLog"
PRINT "##########################################################################################################################################"
SELECT SUBSTRING(s.name,1,50) AS 'DATABASE Name',
b.backup_start_date AS 'Full DB Backup Status',
c.backup_start_date AS 'Differential DB Backup Status',
d.backup_start_date AS 'Transaction Log Backup Status'
FROM MASTER..sysdatabases s
LEFT OUTER JOIN msdb..backupset b
ON s.name = b.database_name
AND b.backup_start_date =
(SELECT MAX(backup_start_date)AS 'Full DB Backup Status'
FROM msdb..backupset
WHERE database_name = b.database_name
AND TYPE = 'D') -- full database backups only, not log backups
LEFT OUTER JOIN msdb..backupset c
ON s.name = c.database_name
AND c.backup_start_date =
(SELECT MAX(backup_start_date)'Differential DB Backup Status'
FROM msdb..backupset
WHERE database_name = c.database_name
AND TYPE = 'I')
LEFT OUTER JOIN msdb..backupset d
ON s.name = d.database_name
AND d.backup_start_date =
(SELECT MAX(backup_start_date)'Transaction Log Backup Status'
FROM msdb..backupset
WHERE database_name = d.database_name
AND TYPE = 'L')
WHERE s.name <>'tempdb'
ORDER BY s.name