SQLskills SQL101: Azure SQL Database Monitoring

As Kimberly blogged about earlier this year, SQLskills has an ongoing initiative to blog about basic topics, which we’re calling SQL101. We’re all blogging about things that we often see done incorrectly, technologies used the wrong way, or where there are many misunderstandings that lead to serious problems. If you want to find all of our SQLskills SQL101 blog posts, check out SQLskills.com/help/SQL101.

If you are ready to start working with Azure SQL Database, which is Microsoft’s Platform as a Service (PaaS) offering for SQL Server, you will want some good diagnostic queries to use to help you better understand how your databases are performing and to give you more information than is exposed by the Azure Portal. After much delay and procrastination on my part, I have finally developed a version of my SQL Server Diagnostic Information Queries that are tailored for use with Azure SQL Database.

Azure SQL Database Diagnostic Information Queries

Azure SQL Database Blank Results Spreadsheet

The basic instructions for using these queries is that you should run each query in the set, one at a time (after reading the directions for that query). It is not really a good idea to simply run the entire batch in one shot, especially the first time you run these queries on a particular server, since some of these queries can take some time to run, depending on your workload and hardware. I also think it is very helpful to run each query, look at the results (and my comments on how to interpret the results) and think about the emerging picture of what is happening on your server as you go through the complete set. I have quite a few comments and links in the script on how to interpret the results after each query.

After running each query, you need to click on the top left square of the results grid in SQL Server Management Studio (SSMS) to select all of the results, and then right-click and select “Copy with Headers” to copy all of the results, including the column headers to the Windows clipboard. Then you paste the results into the matching tab in the blank results spreadsheet.

Please let me know what you think of these queries, and whether you have any suggestions for improvements. Thanks!

SQLskills SQL101: Creating SQL Server Databases

As Kimberly blogged about earlier this year, SQLskills has an ongoing initiative to blog about basic topics, which we’re calling SQL101. We’re all blogging about things that we often see done incorrectly, technologies used the wrong way, or where there are many misunderstandings that lead to serious problems. If you want to find all of our SQLskills SQL101 blog posts, check out SQLskills.com/help/SQL101.

One seemingly simple task that I very often see being done in a less than optimal way is creating a new database in SQL Server. Whether it is done with the SQL Server Management Studio (SSMS) GUI, or with a T-SQL CREATE DATABASE command, many people and organizations are creating new SQL Server databases without really thinking about what they are doing, and without taking advantage of a number of beneficial options and properties.

A SQL Server database requires one data file in the PRIMARY file group and one transaction log file. A very high percentage of SQL Server databases that I see in the wild only have these two required files, which can be problematic for a number of reasons related to both manageability and performance.

You should get in the habit of creating a new file group called MAIN, that is the default file group, that contains two or more data files that are the same size, with the same auto growth increment. If you do this, only the system objects will be in the required data file in the PRIMARY file group, while all of your user objects will be in the other data files in the MAIN file group. This will let you locate your data files across multiple LUNs (either now or in the future), which will make them easier to manage and potentially give you better I/O performance (if those LUNs actually map to separate underlying storage).

When you create a new SQL Server database, it inherits most of its properties from the model system database (unless you explicitly override those properties with ALTER DATABASE commands). By default, SQL Server creates the files for the database in the default location that was specified when SQL Server was installed, unless someone has changed those default locations using the Server Properties: Database Settings dialog shown in Figure 1.

If you do change these database default locations, you should make 100% sure that the new locations actually exist in your file system (since SQL Server does not validate them when you change them). If you change them to a non-existent location, and later try to install a SQL Server Service Pack or Cumulative Update, the Database Engine portion of the installation will fail at the end of the setup process, which could be an unpleasant surprise!

image

Figure 1: Server Properties: Database Settings Dialog

Another way to change the location and properties of your database files is by explicitly specifying what you want when you create the database, or afterwards, with an ALTER DATABASE command.

If you use the SSMS GUI to create a new database as shown in Figure 2, it only requires that you enter a name for the database, and then click the OK button. Even though this will work, it is not really the best method to create a new database. Instead, you should take the time to think about what you are doing and then change a few properties and settings from their default values.

image

Figure 2: New Database: General Dialog with default values

The first thing you should change is the Owner of the database. You should change it from <default> to sa, to ensure that your login is not the owner of the database. Next, you should change the initial size of the files to a more appropriate, larger value. You should also change the Autogrowth increment size for the files to a more appropriate, larger value that is a fixed size in megabytes rather than a percentage-based value. Finally, you may want to change the location where your initial database files will be located.  Your dialog should look something like what you see in Figure 3. After all of this, don’t click OK, because you are not done yet.

 

image

Figure 3: New Database: General Dialog with modified values

Next, you should go to the Options page, as shown in Figure 4, and think about whether you want to change any of your initial database property settings. For example, you might want to change the recovery model, the compatibility level, or possibly other settings depending on your workload or SLA requirements. The point here is to carefully consider your choices and make an explicit choice rather than just blindly accepting all of the default properties

 

image

Figure 4: New Database: Options Dialog with default values

 

Next, we want to go to the Filegroups page, and make some changes. You should add a MAIN file group, and make it the default file group, as you see in Figure 5.

image

Figure 5: New Database: Filegroups Dialog with modified values

The next step is to go back to the General page and add some data files to this new MAIN file group. In Figure 6, I have added two new data files to the MAIN file group, setting their properties to appropriate values. If desired, I could change their locations in the file system. After all of this work, do not click on the OK button! Instead, use the Script dropdown to select “Script Action to New Query Window”, so you can review, edit, and save your database creation T-SQL script.

 

image

Figure 6: New Database: General Dialog with final values

As appropriate for a SQL101-level post, this covers the basic options you should consider when creating a database, as opposed to just typing a database name and clicking OK. If you take the time to do this when you first create the database, you will have a lot more flexibility in the future as your database gets larger.

SQL101: AMD EYPC 7000 Series Processors

As Kimberly blogged about earlier this year, SQLskills has an ongoing initiative to blog about basic topics, which we’re calling SQL101. We’re all blogging about things that we often see done incorrectly, technologies used the wrong way, or where there are many misunderstandings that lead to serious problems. If you want to find all of our SQLskills SQL101 blog posts, check out SQLskills.com/help/SQL101.

On June 20, 2017, AMD officially unveiled its new EPYC 7000 Series processor line for one and two-socket servers. These 14nm processors are based on the same Zen architecture as the recent AMD Ryzen desktop processors, with competitive single-threaded performance, along with very high core counts, memory density, and PCIe 3.0 lane counts.

These processors are a system on a chip (SoC) that includes CPU, memory controller, I/O controller and Server Controller Hub, so that no separate chipset is required. They have up to 32 physical cores per SoC, along with Simultaneous Multithreading (SMT), so you get 64 logical cores per SoC. You also get eight memory channels per socket, which means 16 DDR4 DIMMs per socket. This lets you have up to 2TB of RAM in a one-socket server, and 4TB of RAM in a two-socket server (with 128GB DIMMs). More realistically, you can easily and affordably have 512GB of RAM in a single-socket server with 32GB DIMMs.

You also get 128 PCIe 3.0 lanes per socket, which gives you a lot of total I/O capability. One very nice feature of these processors is that AMD does not cripple the lower-end SKUs when it comes to SMT, memory channels or PCIe 3.0 lanes, which is a big, welcome difference from how Intel does things with their product differentiation.

Another difference from Intel (which I actually don’t like) is that AMD does not have higher base clock speeds in their lower core count SKUs, so the existing strategy to reduce your SQL Server core license costs and also get better single-threaded performance by picking “frequency-optimized” low core count (LCC) processors, such as the Intel Xeon E5-2667 v4 is not going to work the same way.

What you can and should do, is to pick the fastest AMD EPYC SKU available at a given physical core count. For example, there are three EPYC 7000 SKUs that have 32 physical cores, the EPYC 7501, the EPYC 7551, and the EPYC 7601. Since the SQL Server core license cost will be the same, you should pick the EPYC 7601, to get the most performance possible for your license dollars.

Another important caveat is for SQL Server Standard Edition users. SQL Server 2016 Standard Edition has a license limit of four sockets or 24 physical cores, whichever is lower. This means that you need to be very careful what processor you choose for SQL Server 2016 Standard Edition. You basically have five choices to avoid exceeding these license limits.

You can get a 24-core EPYC 7401P, or a 16-core EPYC 7351P for a single-socket server. The 32-core EPYC 7551P would exceed your Standard Edition license limit. In a two-socket server, you can choose an eight-core EPYC 7251, and populate either one or two sockets. You can also choose a 16-core EPYC 7351 or a 24-core EPYC 7451 and just populate one socket of a two-socket server.

If you buy a new database server that has more than 24 physical cores, SQL Server 2016 Standard Edition will only use 24 physical cores per instance, but Microsoft will still expect you to pay for a core license for every physical core present in the machine. This could be a very expensive mistake. In a worst case scenario, you buy a two-socket server with two of the 32-core EPYC 7601 processor, and end up having to pay about $72K in extra license costs for cores that you are not allowed to use in a single instance.

Microsoft has not formally announced any change in these license limits for SQL Server 2017 Standard Edition, but hopefully they will raise these license limits to a more realistic value for modern processors from AMD (and for the upcoming Intel Skylake-SP processors).