SQLintersection Spring 2019

As we head towards our SQLintersection Spring 2019 in June (our 13th show), I’m excited to say that it’s once again our most diverse, complete, and information-packed show yet!

One of the pieces of feedback we hear over and over is that attendees love SQLintersection because it’s a smaller, laid-back show, where you get to actually spend time talking with the presenters 1-1. I have to say that’s one of the reasons why we love the show so much; *we* get to spend time talking to attendees, rather than being mobbed by hundreds of people after a session ends. And we only pick presenters who we know personally, and who we know to be humble, approachable, and eager to help someone out.

We have 2 pre-con days at the show and with our post-con day, there are 8 full-day workshops from which to choose. We have 32 technology-focused (NOT marketing) sessions from which to choose, plus two SQL Server keynotes, multiple industry-wide keynotes by Microsoft executives, and the ever-lively closing Q&A that we record as a RunAs Radio podcast (you can listen to the recording from our Fall 2018 show here.)

You’ll learn proven problem-solving techniques and technologies you can implement immediately. Our focus is around performance monitoring, troubleshooting, designing for scale and performance, cloud, as well as new features in Azure and SQL Server 2016, 2017, and 2019. It’s time to determine your 2008 / 2008 R2 migration strategy – should you upgrade to 2016/2017/2019 directly? Should you move to Azure? This is the place to figure that out!

If you’re interested in how we got here – check out some of Kimberly’s past posts:

  1. SQLintersection: a new year, a new conference
  2. SQLintersection’s Fall Conference – It’s all about ROI!
  3. Fall SQLintersection is coming up soon and we can’t wait!
  4. SQLintersection Conference and SQLafterDark Evening Event – what a fantastic week in Vegas

And Kimberly recorded a Microsoft Channel 9 video where she discusses the 2017 Spring show – see here.

SQLafterDark

With minimal to no marketing filler, we’ve largely kept our conference focus on ROI and technical content (performance / troubleshooting / tales-from-the-trenches with best practices on how to fix them ) but we’ve also added even more social events so that you really get time to intersect with the conference attendees and speakers. The addition of the SQL-specific, pub-quiz-style evening event SQLafterDark was wildly popular from our past shows and that’s returning for Spring!

 

SQLintersection: Great Speakers!

Once again, I think a great show starts with great speakers and current / useful content. All of these speakers are industry-experts that have worked in data / SQL for years (some can even boast decades) but all are still focused on consulting and working in the trenches. And, they’re good presenters! Not only will you hear useful content but you’ll do so in a way that’s digestible and applicable. Every speaker is either an MCM (Master), a past/present SQL Server MVP, or a past/present Microsoft employee (or a combination of all three!) But, regardless of their official credentials, ALL are focused on providing the most ROI that’s possible in their session(s) and/or their workshops, and ALL have spoken for SQLintersection multiple times.

Check out this phenomenal list of speakers:

  • Aaron Bertrand – MVP, SentryOne
  • Ben Miller – MCM, MVP, past Microsoft, Consultant
  • Bob Ward – SQL Server team at Microsoft
  • Brent Ozar – MCM, past-MVP, Consultant
  • Buck Woody – SQL Server team at Microsoft
  • David Pless – MCM, Microsoft
  • Gareth Swanepoel – SQL Server team at Microsoft
  • Glenn Berry – past-MVP, SQLskills
  • Grant Fritchey – MVP, RedGate
  • Jonathan Kehayias – MVP, MCM, MCM Instructor, SQLskills
  • Kevin Farlee – SQL Server team at Microsoft
  • Kimberly L. Tripp – MCM Instructor, MVP, past SQL Server team at Microsoft, SQLskills
  • Pam Lahoud – MCM, SQL Server team at Microsoft
  • Paul S. Randal – MCM Instructor, MVP, past SQL Server team at Microsoft, SQLskills
  • Pedro Lopes – MCM, SQL Server team at Microsoft
  • Tim Chapman – MCM, Microsoft
  • Tim Radney – MVP, SQLskills

You can read everyone’s full bio on our speaker page here.

SQLintersection: When is it all happening?

The conference officially runs from Tuesday, June 11 through Thursday, June 13 with pre-conference and post-conference workshops that extend the show over a total of up to 6 full days. For the full conference, you’ll want to be there from Sunday, June 9 through Friday, June 14.

  • Sunday, June 9 – pre-con day. There are three workshops running:
    • Migrating to SQL Server 2019 with Glenn Berry
    • Developer’s SQL Server Recipe Book with Brent Ozar
    • Leveling up with PowerShell for the DBA with Ben Miller
  • Monday, June 10 – pre-con day. There are three workshops running:
    • Performance Tuning and Optimization for Modern Workloads with Tim Radney and David Pless
    • Performance Troubleshooting using Waits and Latches with Paul S. Randal
    • SQL Server Indexes: What, Why, and How! with Kimberly L. Tripp
  • Tuesday, June 11 through Thursday, June 13 is the main conference. Conference sessions will run all day in multiple tracks:
    • Check out our sessions online here
    • Be sure to check out our cross-conference events and sessions
    • Get your pop-culture trivia and techie-SQL-trivia hat on and join us for SQLafterDark on Wednesday evening, June 12
  • Friday, June 14 is our final day with three post-conference workshops running:
    • SQL Server Reporting Services and Power BI – Reporting Solutions with David Pless and Tim Radney
    • Zero to Hero: Faster SQL Query Performance with Jonathan Kehayias

SQLintersection: Why is it for you?

If you want practical information delivered by speakers that not-only know the technologies but are competent and consistently highly-rated presenters – this is the show for you. You will understand the RIGHT features to troubleshoot and solve your performance and availability problems now!

Check us out: www.SQLintersection.com.

We hope to see you there!

PS – Use the discount code ‘SQLskills’ when you register and receive $50 off registration!

The Curious Case of… unstoppable DBCC CHECKDB

(The Curious Case of… used to be part of our bi-weekly newsletter but we decided to make it a regular blog post instead so it can sometimes be more frequent. It covers something interesting one of us encountered when working with a client, doing some testing, or were asked in a random question from the community.)

Last week I was asked to explain an unstoppable DBCC CHECKDB process. The client regularly runs DBCC CHECKDB at night during off-peak load times and has no problems with it but last week they ran it during the day because of a corruption indication, then decided to stop it and couldn’t. They killed the spid and nothing happened except the spid showed as being in the KILLED/ROLLBACK state.

Let’s back up a bit and explain what DBCC CHECKDB is doing under the covers.

It needs to see a transactionally-consistent, unchanging view of the database. Prior to SQL Server 2005 it did this by essentially running its own version of crash recovery inside itself, by analyzing the database’s transaction log. This was pretty tortuous code that I helped write for SQL Server 2000, there were some cases where it caused false positives, and I had great fun one week in 2001 or 2002 removing all that code forever. The replacement for that code was to instead use a private database snapshot, because a database snapshot gives a transactionally-consistent, unchanging view of a database.

So, the first thing DBCC CHECKDB does in SQL Server 2005 and later is create a database snapshot (unless you specified WITH TABLOCK, or the target database is read-only, single-user, or already a database snapshot).

A database snapshot runs crash recovery of the target database *into* the database snapshot, and herein lies the problem: crash recovery cannot be interrupted, and there’s no check in the crash recovery code to tell whether it’s *real* crash recovery, or crash recovery into a database snapshot (or into the DBCC CHECKDB private snapshot).

If there’s a lot of transaction log to be recovered as part of the initial crash recovery into the database snapshot, that could take a long time. And if someone tries to kill the DBCC CHECKDB while the database snapshot is still running crash recovery, nothing will happen until the crash recovery finishes. It’s not a DBCC shortcoming, it’s a database snapshot shortcoming, and you just have to let it finish.

So there you go – mystery explained!

The Curious Case of… faster disks causing more WRITELOG waits

(The Curious Case of… used to be part of our bi-weekly newsletter but we decided to make it a regular blog post instead so it can sometimes be more frequent. It covers something interesting one of us encountered when working with a client, doing some testing, or were asked in a random question from the community.)

I was contacted last week by someone who was asking how are faster disks causing more WRITELOG waits. They were seeing lots of these waits, with an average wait time of 18ms. The log was stored on a Raid-1 array, using locally-attached spinning disks in the server. They figured that by moving the log to Raid-1 array of SSDs, they’d reduce the WRITELOG wait time and get better workload throughput.

They did so and got better performance, but were very surprised to now see WRITELOG as the most frequent wait type on the server, even though the average wait time was less than 1ms, and asked me to explain.

Let me start by saying this is not unusual, and I’ve seen this sequence of events many times with clients and students.

WRITELOG wait occurs when a log block is being flushed to disk, and the two most common causes of this are that the log block is full (60KB) or a transaction committed. If it’s the case that a workload is comprised of lots of very small transactions, then a reduction in wait time for the log block flush from 18ms down to less than 1ms is very likely to result in a higher throughput. This means more transactions happening and hence more log block flushes when the transactions commit, so *more* WRITELOG waits.

Bottom line: What was confusing to them was that performance tuning is usually about reducing the incidence of waits and/or the individual wait times. This is about the only example I can think of where you do something that increases performance and the wait type you were concerned about becomes more prevalent.