Scheduler Stories

Ewald Cress has a couple of posts about the scheduler.  First, fiber mode scheduling:

The title of this post is of course an allusion to Ken Henderson’s classic article The perils of fiber mode, where he hammers home the point that fiber scheduling, a.k.a. lightweight pooling, appears seductive until you realise what you have to give up to use it.

We’ll get to the juicy detail in a moment, but as a reminder, the perils of fibers lie in their promiscuity: many fibers may share one thread, its kernel structures and its thread-local storage. This is no problem for code that was written with fibers in mind, including all of SQLOS, but unfortunately there are bodies of code for which this isn’t true.

Next up is the Windows scheduler:

Hardware interrupts, which run in kernel mode and return to user mode quickly, should be nothing more than tiny hiccups in a running thread’s quantum. The other 90% of the interrupt iceberg manifests in user mode as Deferred Procedure Calls (DPCs, or “bottom halves” to the Linux crowd) but should still only steal small change in terms of CPU cycles. Context switches to another thread represent a completely different story, because it could be ages before control returns to our thread, meaning that our fiber scheduler is completely out of commission for a while.

This possibility – a SQLOS scheduler losing the CPU for an extended period – is just one of those things we need to live with, but on a sane server, it shouldn’t be something to be too concerned about. Consider that this happens all the time in virtualised environments, where our vCPU can essentially cease to exist while another VM has a ride on the physical CPU.

These are fairly long reads, but we’re getting to levels where you can see these settings in the Database Engine (like Lightweight Pooling).

Query Store Filegroups

Kendra Little links to a Connect item:

Can you change the filegroup where Query Store keeps its data?

I thought there might be a trick to use a different filegroup for Query Store by using the default filegroup setting in SQL Server before enabling it, but NOPE!

Please vote for this to be improved in this Connect Item.

I concur; Query Store can grow to be pretty large on busy systems, so diligent DBAs who want to keep PRIMARY as small as possible will suddenly find a multi-gigabyte Query Store slowing down those PRIMARY filegroup restores.

Using SQL Server DMA

Arun Sirpal has a series on using the new Data Migration Assistant.  Part 1:

It “enables you to upgrade to a modern data platform by detecting compatibility issues that can impact database functionality on your new version of SQL Server. It recommends performance and reliability improvements for your target environment. It allows you to not only move your schema and data, but also uncontained objects from your source server to your target server”. It can be found at this link:https://www.microsoft.com/en-us/download/details.aspx?id=53595.

Part 2 is all about performing a migration:

By the way the backup file created via the tool is temporary, after a migration it is deleted. Also the compatibility level DOES NOT change, you need to do this yourself.

I haven’t used this tool yet, but it does look like an upgrade to the old Upgrade Advisor.

Thinking About Compile Time

Jay Robinson has a post on compilation time and especially indexed views:

What I found was that worker time needed to compile these queries is indistinguishable from that needed to execute them. To show this, let’s look at an example in AdventureWorks2014. In this example, I’m going to create and execute two similar procedures. I’m also going to create a number of indexed views.

Why indexed views? I want to increase compile time significantly for this exercise, and a large number of indexed views can do that. From MSDN: “The query optimizer may use indexed views to speed up the query execution. The view does not have to be referenced in the query for the optimizer to consider that view for a substitution.” My thanks tooas_public on stackoverflow.com for that tip.

Indexed views come at a cost, as Jay shows.

Azure SQL DW Statistics

Emma Stewart looks at how statistics are created in Azure SQL Data Warehouse:

In Azure SQL Data Warehouse, statistics have to be created manually. On previous SQL Server projects, creating and maintaining statistics wasn’t something that we had to incorporate into our design (and really think about!) however with SQL DW we need to make sure we think about how to include it in our process in order to make sure we take advantage of the benefits of working with Azure DW.

The major selling point of Azure SQL Data Warehouse is that it is capable of processing huge volumes of data, one of the specific performance optimisations that has been made is the distributed query optimiser. Using the information obtained from the statistics (information on data size and distribution), the service is able to optimize queries by assessing the cost of specific distributed query operations. Therefore, since the query optimiser is cost-based, SQL DW will always choose the plan with the lowest cost.

Azure SQL Data Warehouse is a bit of a strange animal, with differences in statistics being one of the smaller changes versus “classic” SQL Server.

Details On Azure SSAS

James Serra breaks down what Azure Analysis Services has to offer:

  • Developers can use SQL Server Data Tools (SSDT) in Visual Studio for creating models and deploying them to the service.  Administrators can manage the models using SQL Server Management Studio (SSMS) and investigate issues using SQL Server Profiler

  • Business users can consume the models in any major BI tool.  Supported Microsoft tools include Power BI, Excel, and SQL Server Reporting Services.  Other MDX compliant BI tools can also be used, after downloading and installing the latest drivers

  • The service currently supports tabular models (compatibility level 1200 only).  Support for multidimensional models will be considered for a future release, based on customer demand

Between tabular-only support and the max size being 100 GB (if I’m reading this correctly), they’re not yet ready to push the product hard.  Given that it just came out, that makes sense, and hopefully the training wheels come off.

Sparklines

Devin Knight continues his custom visuals series with the sparkline:

Key Takeaways

  • Shows trends in data most often by time.

  • It can only visualize a single measure.

  • The Sparkline can be configured to animate across time.

In the right circumstances, I love sparklines.  My circumstances are as follows:  when you are viewing time series data for relatively few elements in which the trend is more important than the levels.  In that scenario, sparklines are efficient and tell the story without extraneous numbers or clutter getting in the way.

Latency Vs Throughput

I have a post up on understanding latency versus throughput:

The primary method in which data moves from one process to another is through buffers.  We break up data into smaller portions and push them to their destination.  In Integration Services, we have buffers.  When passing data through TCP, we use packets.

Okay, so what’s the trade-off?  The trade-off is between latency and throughput.  Let’s take TCP packets as an example.  Say you have a series of 50-byte messages you want to send from a source to a destination.  We have two primary options:  push messages as fast as possible, or hold off until you have the most data you can store in a packet and send it along.  For simplicity’s sake, we’ll say that we can fit about 1350 bytes in a packet, so we can store 27 messages in a packet.  We’ll also assume that it takes 10 milliseconds to send a packet from the source to the destination (regardless of packet size, as we’re using powerful connections) and 1 millisecond to produce a message.

We use pipes as metaphors in IT, especially around data transfer, and I think it’s a solid metaphor because it intuitively includes most of the important concepts we need to worry about with data.  We have latency (how long it takes something to go from one end of the pipe to the other), throughput (how much we can move at any point in time, which is determined by things like the diameter of the pipe), back pressure (in the pipe scenario, resistance caused by pipe directional changes; in the data world, when downstream operators are slower than upstream operators), etc.

Visualizing Checkins

Rob Sewell uses Power BI to map where he’s been:

I am using the swarm API but the principle is the same for any other API that provides you with data. For example, I used the same principles to create the embedded reports on the PASS PowerShell Virtual Chapter page showing the status of the cards suggesting improvements to the sqlserver module for the product team to work on. Hopefully, this post will give you some ideas to work on and show you that it is quite easy to get excellent data visualisation from APIs

First up we need to get the data. I took a look at the Swarm developers page ( The Trello ishere by the way) I had to register for an app, which gave me a client id and a secret. I then followed the steps here to get my user token I was only interested in my own check ins so I used the steps under Token flow Client applications to get my access token which I used in an URL like this.

This post includes some Powershell and quite a few animated GIFs, making it easy to follow.

View Performance

Grant Fritchey looks at view performance vis-a-vis stored procedures:

The difference in the performance including compile time for the procedure alone is 700mc better on average than the view. That’s an 8% difference. It was almost that high for the view that used the procedure at 7%.

If we’re just talking compile time then, there is a significant win if we avoid the view. This is no doubt because of the extra work involved in unpacking the view and going through the simplification process within the optimizer. Plus, the view alone in our query was parameterized by the optimizer in order to assist it’s performance over time (as we saw in the average results without the recompile). All that extra work explains the 8% difference.

Read the whole thing.

Categories

November 2016
MTWTFSS
« Oct Dec »
 123456
78910111213
14151617181920
21222324252627
282930