Pass Summit 2016

PASS summit 2016 took place (as usually these last years) in Seattle, Washington. For those who don’t know, it is the world’s largest and most intensive technical training conference for Microsoft SQL Server and BI professionals it is planned and presented by the SQL Server community for the SQL Server community.
This year, the summit attracted more than 4000 attendees from all around the world for one week of learning, sharing and networking. Tons of technical sessions, community zones and an effective organization have built the success of this year’s event.
PASS Summit themes went definitely beyond SQL Server data platform and Business intelligence with great talks about Cloud computing in Azure, continuous integration, machine learning in the Microsoft environment.
The first two days were dedicated to pre-conference sessions : Basically a deep dive in one topic during all the day with the same speaker, pretty interesting to learn something new or to fine-tune some skills.
During the next three days, a plenty of technical talks were scheduled, all the tracks were addressed, enough to satisfy all the trades in our opinion.

This year, we were 3 Criteo’s DBA that had the chance to fly to Seattle to take part to this awesome event. We lost no time to get into the swing of this event. We attended many good sessions especially these 3 below.

SQL Server on Linux by Tobias Tesrnstrom, Slava Oks

Back in last March, Microsoft announced SQL Server on Linux, Today, there’s even a public preview available for those who couldn’t believe that could happen one day.
We’re talking here about the next edition of SQL Server (aka. vNext), or the edition after SQL Server 2016.
Obviously, all the features are not yet available on this future edition, but guess what, the speakers who were part of SQL Server Development and Product team at Microsoft shared their roadmap and explained that almost all features will be increasingly enabled (except FileTable).
The question arises as to how the hell did they do to get rid of Win32 API dependence in SQLOS ?
Actually SQL Servers runs on Linux by using SQLPAL (SQL platform abstraction layer). Inside the SQLPAL, we find SQLOSv2, the new version of SQLOS that communicates either with the former SQLOS Api or with the new Win32 like APIs. The windows clustering foundation layer will be replaced by Linux Pacemaker (http://wiki.clusterlabs.org/wiki/Pacemaker) to bring clustering capabilities (Failover clustering,
AlwaysOn availability groups) to SQL on Linux.
Nowadays, SQL Server runs either on Ubuntu or RHEL Linux. It’s even possible to run SQL Server in a docker container (https://hub.docker.com/r/microsoft/mssql-server-windows/).
There’s no doubt, we’re looking forward at Criteo to give this vNext edition of SQL Server a try…and maybe more!

Why your data type choices matter ? by Andy Yun

At Criteo, we always review T-SQL code. At least 2 DBA check every SQL Server commits. So one of our daily tasks is to help developers, to give them advice on how to build a good data model.
In this session, in the developer track, Andy explained how the engine stores data and why it is important to chose correctly datatype.
As a DBA, I’m not sure I improved my technical skills but I’m sure this session will help me to explain easily to developers which datatype to chose.
After the session, I continued to speak with Andy and he gave me a good tip about how to detect implicit conversion from plan cache. (https://www.sqlskills.com/blogs/jonathan/finding-implicit-column-conversions-in-the-plan-cache/)

Intro to internals, How to think like the SQL Server engine by Brent Ozar

I went to this session since the topic looked really similar to one of the session we provide to our developers to show them how things works and is processed when running a query.
As I thought, the session dealt with very basic concepts on how SQL Server builds execution plans, what are clustered and nonclustered indexes, how they help reduce IO and CPU usage, the way SQL Server caches data pages, why statistics are important and how do they work in cardinality estimation.
Brent Ozar is a really good speaker, and the session was really good and really well explained in a very humoristic manner too, it gave me ideas on how to manage futures sessions.

 

We could speak many hours about these sessions. Some of them were very interesting :

  • Service Broker by Ed Leighton-Dick
  • Inside SQL Server In-Memory OLTP Bob Ward
  • The second Keynote by David DeWitt (excellent!)
  • Hacking SQL Server for fun and profit by Argenis Fernandez
  • ColumnStore index improvements by Niko Neugebauer
  • From FearOps to Devops by Hamish Watson
  • R services integration in SQL Server 2016 by Umachandar Jayachandran

 

On thursday evening, we organised a dinner with all French and Canadian we met during the week. We were 11 French data professional in a Steakhouse. A very good moment.

We’d really like to thanks Criteo for this great trip  (sourire)