SQL Server Architecture

MS SQL Server is a client-server architecture. MS SQL Server process starts with the client application sending a request. The SQL Server accepts, processes and replies to the request with processed data. Let's discuss in detail the entire architecture shown below:

As the below Diagram depicts there are three major components in SQL Server Architecture:

1. Protocol Layer

2. Relational Engine

3. Storage Engine

SQL Server Architecture Diagram

Let's discuss in detail about all the three above major modules. In this tutorial, you will learn.

• Protocol Layer - SNI
• Shared Memory
• TCP/IP
• Named Pipes
• What is TDS?
• Relational Engine
• CMD Parser
• Optimizer
• Query Executor
• Storage Engine
• File types
• Access Method
• Buffer Manager
• Plan Cache
• Data Parsing: Buffer cache & Data Storage
• Transaction Manager

Protocol Layer - SNI

MS SQL SERVER PROTOCOL LAYER supports 3 Type of Client Server Architecture. We will start with "Three Type of Client Server Architecture" which MS SQL Server supports.

Shared Memory

Let's reconsider an early morning Conversation scenario.

MOM and TOM - Here Tom and his Mom, were at the same logical place, i.e. at their home. Tom was able to ask for Coffee and Mom was able it serve it hot.

MS SQL SERVER - Here MS SQL server provides SHARED MEMORY PROTOCOL. Here CLIENT and MS SQL server run on the same machine. Both can communicate via Shared Memory protocol.

Analogy: Lets map entities in the above two scenarios. We can easily map Tom to Client, Mom to SQL server, Home to Machine, and Verbal Communication to Shared Memory Protocol.

TCP/IP

Now consider in the evening, Tom is in the party mood. He wants a Coffee ordered from a well-known Coffee Shop. The Coffee shop is located 10 km away from his home.

Here Tom and Starbuck are in different physical location. Tom at home and Starbucks at the busy marketplace. They're communicating via Cellular network. Similarly, MS SQL SERVER provides the capability to interact via TCP/IP protocol, where CLIENT and MS SQL Server are remote to each other and installed on a separate machine.

Analogy: Lets map entities in the above two scenarios. We can easily map Tom to Client, Starbuck to SQL server, the Home/Market place to Remote location and finally Cellular network to TCP/IP protocol.

Notes from the desk of Configuration/installation:

• In SQL Management Studio - For Connection via TCP\IP, "Server Name" Option has to be "Machine\Instance of the server."
• SQL server uses port 1433 in TCP/IP.

Named Pipes

Now finally at night, Tom wanted to have a light green tea which her neighbor, Sierra prepare very well.

Here Tom and his Neighbor, Sierra, are in same physical location, being each other's neighbor. They're communicating via Intra network. Similarly, MS SQL SERVER provides the capability to interact via the Named Pipe protocol. Here the CLIENT and MS SQL SERVER are in connection via LAN.

Analogy: Lets map entities in the above two scenarios. We can easily map Tom to Client, Sierra to SQL server, Neighbor to LAN and finally Intra network to Named Pipe Protocol.

Notes from the desk of Configuration/installation:

• For Connection via Named Pipe. This option is disabled by default and needs to be enabled by the SQL Configuration Manager.

What is TDS?

Now that we know that there are three types of Client-Server Architecture, lets us have a glance at TDS:

• TDS stands for Tabular Data Stream.
• All 3 protocols use TDS packets. TDS is encapsulated in Network packets. This enables data transfer from the client machine to the server machine.
• TDS was first developed by Sybase and is now Owned by Microsoft

Relational Engine

The Relational Engine is also known as the Query Processor. It has the SQL Server components that determine what exactly a query needs to do and how it can be done best. It is responsible for the execution of user queries by requesting data from the storage engine and processing the results that are returned.

As depicted in the Architectural Diagram there are 3 major components of the Relational Engine. Let's study the components in detail:

CMD Parser

Data once received from Protocol Layer is then passed to Relational Engine. "CMD Parser" is the first component of Relational Engine to receive the Query data. The principal job of CMD Parser is to check the query for Syntactic and Semantic error. Finally, it generates a Query Tree. Let's discuss in detail.

Syntactic check:

• Like every other Programming language, MS SQL also has the predefined set of Keywords. Also, SQL Server has its own grammar which SQL server understands.
• SELECT, INSERT, UPDATE, and many others belong to MS SQL predefined Keyword lists.
• CMD Parser does syntactic check. If users' input does not follow these language syntax or grammar rules, it returns an error.

Example: Let's say a Russian went to a Japanese restaurant. He orders fast food in the Russian language. Unfortunately, the waiter only understands Japanese. What would be the most obvious result?

The Answer is – the waiter is unable to process the order further.

There should not be any deviation in Grammar or language which SQL server accepts. If there are, SQL server cannot process it and hence will return an error message.

We will learn about MS SQL query more in upcoming tutorials. Yet, consider below most basic Query Syntax as

SELECT * from <TABLE_NAME>;

Now, to get the perception of what syntactic does, say if the user runs the basic query as below:

SELECR * from <TABLE_NAME>

Note that instead of 'SELECT' user typed "SELECR."

Result: THE CMD Parser will parse this statement and will throw the error message. As "SELECR" does not follow the predefined keyword name and grammar. Here CMD Parser was expecting "SELECT."

Semantic check:

• This is performed by Normalizer.
• In its simplest form, it checks whether Column name, Table name being queried exist in Schema. And if it exists, bind it to Query. This is also known as Binding.
• Complexity increases when user queries contain VIEW. Normalizer performs the replacement with the internally stored view definition and much more.

let's understand this with help of below example -

SELECT * from USER_ID

Result: THE CMD Parser will parse this statement for Semantic check. The parser will throw an error message as Normalizer will not find the requested table (USER_ID) as it does not exist.

Create Query Tree:

• This step generates different execution tree in which query can be run.
• Note that, all the different trees have the same desired output.

Optimizer

The work of the optimizer is to create an execution plan for the user's query. This is the plan that will determine how the user query will be executed.

Note that not all queries are optimized. Optimization is done for DML (Data Modification Language) commands like SELECT, INSERT, DELETE, and UPDATE. Such queries are first marked then send to the optimizer. DDL commands like CREATE and ALTER are not optimized, but they are instead compiled into an internal form. The query cost is calculated based on factors like CPU usage, Memory usage, and Input/ Output needs.

Optimizer's role is to find the cheapest, not the best, cost-effective execution plan.

Before we Jump into more technical detail of Optimizer consider below real-life example:

Example:

Let's say, you want to open an online Bank account. You already know about one Bank which takes a maximum of 2 Days to open an account. But, you also have a list of 20 other banks, which may or may not take less than 2 days. You can start engaging with these banks to determine which banks take less than 2 days. Now, you may not find a bank which takes less than 2 Days, and there is additional time lost due to the search activity itself. It would have been better to open an account with the first bank itself.

Conclusion: It's is more important to select wisely. To be precise, choose which option is best, not the cheapest.

Similarly, MS SQL Optimizer works on inbuilt exhaustive/heuristic algorithms. The goal is to minimize query run time. All the Optimizer algorithms are propriety of Microsoft and a secret. Although, below are the high-level steps performed by MS SQL Optimizer. Searches of Optimization follows three phases as shown in the below diagram:

Phase 0: Search for Trivial Plan:

• This is also known as Pre-optimization stage.
• For some cases, there could be only one practical, workable plan, known as a trivial plan. There is no need for creating an optimized plan. The reason is, searching more would result in finding the same run time execution plan. That too with the extra cost of Searching for optimized Plan which was not required at all.
• If no Trivial plan found, then 1^st Phase starts.

Phase 1: Search for Transaction processing plans

• This includes the search for Simple and Complex Plan.
• Simple Plan Search: Past Data of column and Index involved in Query, will be used for Statistical Analysis. This usually consists but not restricted to one Index Per table.
• Still, if the simple plan is not found, then more complex Plan is searched. It involves Multiple Index per table.

Phase 2: Parallel Processing and Optimization.

• If none of the above strategies work, Optimizer searches for Parallel Processing possibilities. This depends on the Machine's processing capabilities and configuration.
• If that is still not possible, then the final optimization phase starts. Now, the final optimization aim is finding all other possible options for executing the query in the best way. Final optimization phase Algorithms are Microsoft Propriety.

Query Executor

Query executer calls Access Method. It provides an execution plan for data fetching logic required for execution. Once data is received from Storage Engine, the result gets published to the Protocol layer. Finally, data is sent to the end user.

Storage Engine

The work of the Storage Engine is to store data in a storage system like Disk or SAN and retrieve the data when needed. Before we deep dive into Storage engine, let's have a look at how data is stored in Database and type of files available.

Data File and Extent:

Data File, physically stores data in the form of data pages, with each data page having a size of 8KB, forming the smallest storage unit in SQL Server. These data pages are logically grouped to form extents. No object is assigned a page in SQL Server.

The maintenance of the object is done via extents. The page has a section called the Page Header with a size of 96 bytes, carrying the metadata information about the page like the Page Type, Page Number, Size of Used Space, Size of Free Space, and Pointer to the next page and previous page, etc.

File types

Access Method

It acts as an interface between query executor and Buffer Manager/Transaction Logs.

Access Method itself does not do any execution.

The first action is to determine whether the query is:

1. Select Statement (DDL)

2. Non- Select Statement (DDL & DML)

Depending upon the result, the Access Method takes the following steps:

1. If the query is DDL, SELECT statement, the query is pass to the Buffer Manager for further processing.

2. And if query if DDL, NON-SELECT statement, the query is pass to Transaction Manager. This mostly includes the UPDATE statement.

Buffer Manager

Buffer manager manages core functions for modules below:

• Plan Cache
• Data Parsing: Buffer cache & Data storage
• Dirty Page

We will learn Plan, Buffer and Data cache in this section. We will cover Dirty pages in the Transaction section.

Plan Cache

• Existing Query plan: The buffer manager checks if the execution plan is there in the stored Plan Cache. If Yes, then query plan cache and its associated data cache is used.
• First time Cache plan: Where does existing Plan cache come from?

If the first-time query execution plan is being run and is complex, it makes sense to store it in in the Plane cache. This will ensure faster availability when the next time SQL server gets the same query. So, it's nothing else but the query itself which Plan execution is being stored if it is being run for the first time.

Data Parsing: Buffer cache & Data Storage

Buffer manager provides access to the data required. Below two approaches are possible depending upon whether data exist in the data cache or not:

Buffer Cache - Soft Parsing:

Buffer Manager looks for Data in Buffer in Data cache. If present, then this Data is used by Query Executor. This improves the performance as the number of I/O operation is reduced when fetching data from the cache as compared to fetching data from Data storage.

Data Storage - Hard Parsing:

If data is not present in Buffer Manager than required Data is searched in Data Storage. If also stores data in the data cache for future use.

Dirty Page

It is stored as a processing logic of Transaction Manager. We will learn in detail in Transaction Manager section.

Transaction Manager

Transaction Manager is invoked when access method determines that Query is a Non-Select statement.

Log Manager

• Log Manager keeps a track of all updates done in the system via logs in Transaction Logs.
• Logs have Logs Sequence Number with the Transaction ID and Data Modification Record.
• This is used for keeping track of Transaction Committed and Transaction Rollback.

Lock Manager

• During Transaction, the associated data in Data Storage is in the Lock state. This process is handled by Lock Manager.
• This process ensures data consistency and isolation. Also known as ACID properties.

Execution Process

• Log Manager start logging and Lock Manager locks the associated data.
• Data's copy is maintained in the Buffer cache.
• Copy of data supposed to be updated is maintained in Log buffer and all the events updates data in Data buffer.
• Pages which store the data is also known as Dirty Pages.
• Checkpoint and Write-Ahead Logging: This process run and mark all the page from Dirty Pages to Disk, but the page remains in the cache. Frequency is approximately 1 run per minute.But the page is first pushed to Data page of the log file from Buffer log. This is known as Write Ahead Logging.
• Lazy Writer: The Dirty page can remain in memory. When SQL server observes a huge load and Buffer memory is needed for a new transaction, it frees up Dirty Pages from the cache. It operates on LRU – Least recently used Algorithm for cleaning page from buffer pool to disk.

Summary:

• Three Type of Client Server Architecture exist: 1) Shared Memory 2) TCP/IP 3)Named Pipes
• TDS, developed by Sybase and now owned by Microsoft, is a packet which is encapsulated in Network packets for data transfer from the client machine to the server machine.
• Relational Engine contains three major components:

CMD Parser: This is responsible for Syntactic and Semantic error & finally generate a Query Tree.

Optimizer: Optimizer role is to find the cheapest, not the best, cost-effective execution plan.

Query Executor: Query executer calls Access Method and provides execution plan for data fetching logic required for execution.

• Three type of files exists Primary file, Secondary file, and Log files.

• Storage Engine: Has following important components

Access Method: This Component Determine whether the query is Select or Non-Select Statement. Invokes Buffer and Transfer Manager accordingly.

Buffer Manager: Buffer manager manages core functions for Plan Cache, Data Parsing & Dirty Page.

Transaction Manager: It manager Non-Select Transaction with help of Log and Lock Managers. Also, facilitates important implementation of Write Ahead logging and Lazy writers.

WAL-SQL Server needs to guarantee the durability of your transactions (once you commit your data it is there even in the event of power loss) and the ability to roll back the data changed from uncommitted transactions. The mechanism that is being utilized is called Write-Ahead Logging (WAL). It simply means that SQL Server needs to write the log records associated with a particular modification before it writes the page to the disk regardless if this happening due to a Checkpoint process or as part of Lazy Writer activity.

Checkpoint is an internal process that writes all dirty pages (modified pages) from Buffer Cache to Physical disk, apart from this it also writes the log records from log buffer to physical file. Writing of Dirty pages from buffer cache to data file is also known as Hardening of dirty pages.

It is a dedicated process and runs automatically by SQL Server at specific intervals. SQL Server runs checkpoint process for each Database individually.

Checkpoint helps to reduce the recovery time for SQL Server in the event of unexpected shutdown or system crash\Failure.

In SQL Server 2012 there are four types of Checkpoints:

Automatic: This is the most common checkpoint which runs as a process in the background to make sure SQL Server Database can be recovered in the time limit defined by the Recovery Interval – Server Configuration Option.

Indirect: This is new in SQL Server 2012. This also runs in the background but to meet a user-specified target recovery time for the specific Database where the option has been configured. Once the Target_Recovery_Time for a given database has been selected this will override the Recovery Interval specified for the server and avoid Automatic Checkpoint on such DB.

Manual: This one runs just like any other T-SQL statement, once you issue checkpoint command it will run to its completion. Manual Checkpoint runs for your current Database Only. You can also specify the Checkpoint Duration which is optional, this duration specifies the time in which you want your checkpoint to complete.

Internal-1. Shutdown initiates a Checkpoint operation on all databases except when Shutdown is not clean (Shutdown with nowait)
2. If the recovery model gets changed from Full\Bulk-logged to Simple.
3. While taking Backup of the Database.
4. If your DB is in Simple Recovery model, checkpoint process executes automatically either when the log becomes 70% full, or based on Server option-Recovery Interval.
5. Alter Database command to add or remove a data\log file also initiates a checkpoint.
6. Checkpoint also takes place when the recovery model of the DB is Bulk-Logged, and a minimally logged operation is performed.
7. DB Snapshot creation.

LAZY WRITER purpose is to release the buffer pool memory (for pages cached in buffer pool) when memory pressure occurs. When more memory is needed (e.g. for bringing in new pages to the cache), lazy writer responds to a memory pressure releasing the “coldest” pages from the buffer pool, and makes more memory available for new pages to come in.

2) Troubleshoot Slow-Running Queries in SQL Server

Blocking issue

Deadlock issue

DB Growth issue

To check CPU and memory uses

To check last UPDATE STATISTIC details

Verify fragmentation level of index

Use execution plain to check index scan, index seek and key lookup

Using SQL Server Profiler

SQL Server Extended Events

The SQL Profiler has been replaced by SQL Server Extended Events. This is sure to anger a lot of people, but I can understand why Microsoft is doing it.

Extended Events works via Event Tracing (ETW). This has been the common way for all Microsoft related technologies to expose diagnostic data.

ETW provides much more flexibility. As a developer, I could easily tap into ETW events from SQL Server to collect data for custom uses. That is really cool and really powerful.

3) What you will to improve SQL Server performance.

To create multiple tempdb file

Keep Data and log file in different drive

Keep system db and user db in separate drive

Assign memory 10 to 90 % to SQL

Rebuild index and update statistic at least every Week

4) What is Covering index? Where can I get benefited by Index?

A covering index is one which can satisfy all requested columns in a query without performing a further lookup into the clustered in

 Index with INCLUDE clause is the covering index.

                      Covering index is basically used to cover the query (include columns from Select list which are not part of index) and to avoid bookmark lookup.

                     1023 columns can add in the INCLUDE clause.

5) I have used ALTER INDEX REBUILD to remove index fragmentation. In some cases, REBUILD does not seem to remove this fragmentation. What are the reasons why REBUILD does not remove fragmentation? It seems that this happens especially with small indices.

This can also happen with very LARGE indexes.

I had some indexes on a table with around 700m rows that I couldn't defragment below around 30%. The issue was not enough contiguous free space inside the database to arrange the index consecutively.

To work around a very large index that won't defragment, the BEST solution is to pre-size a new database and move all your objects to that DB, then recreate your indexes there.

If an index is very small (I believe less than 8 pages) it will use mixed extents. Therefore, it'll appear as if there is still fragmentation remaining, as the housing extent will contain pages from multiple indexes.

Because of this, and also the fact that in such a small index that fragmentation is typically negligible, you really should only be rebuilding indexes with a certain page threshold. It is best practices to rebuild fragmented indexes that are a minimum of 1000 pages.

6) How do I change MAXDOP using SSMS or T-SQL?

Below shows the T-SQL command that you can run to change the value.  In this example I am changing the value to 4. This means if a query uses a parallel execution plan it will only use four of the available processors.

Within SSMS, right click on a registered server and select Property.  Then go to the Advanced page as shown below. You can then change this value and click OK to save the value.

After making this change the value goes into affect immediately, there is not a need to restart SQL Server.

7) FOUR PHASE SQL SERVER DATABASE RECOVERY PROCESS

Following the data copy phase involving copying of all the data, log, and index-pages from the backup media of a database to the database files, four consecutive phases that take place during SQL Server recovery are

1.       Discovery – is to find the logical structure of the Transaction log file.

2.       Analysis – is to find the best LSN starting from which rolling forward can be done during redo phase.

3.       Redo – is the phase during which the changes caused by active transactions (at the time of crash) are hardened onto Data files.

4.       Undo – is the phase where in, rolling back of the active transactions for consistency, takes place.