In this blog we will be covering all the news and announcements from the 2024 PASS Data Summit Keynote as live as we can get it captured. (And later will refine as much as we can, as capturing the details as they happen can be quite tough!)

Intro

Steve Jones from Redgate welcomes all of us to the 2024 PASS Summit! There is a lot of excitement in the air!

A quick thank you to the sponsors 

Save the Date People! 

Next year’s PASS Data Summit will be November 17-21, 2025, right back here in Seattle WA!

Microsoft Keynote: Fuel AI innovation with Azure Databases

Involved with the keynote is a host of people that the bloggers who contributed to this presentation admire (and in some cases, know!):

 Micrsoft Takes the Stage

Shireesh Thota (Corporate Vice President, Azure Databases, Microsoft ) takes the stage.

Shireesh is once again leading the conversation for the Keynote.

The scholarship program is an excellent way to expand the conference to include more people eager to learn. Great program this year!

It is also noted that at this year’s summit there is a concurrent conference on PostgreSQL occurring alongside the PASS Summit. More opportunities for us to learn about more things!

Azure Data Communities

Lots of great communities sharing the knowledge!

SQL Server 2022 Momentum and Commitment

Shireesh reconfirms a commitment to SQL Server product as on on-premises product. Considered as a growth platform!

SSMS 21 with Copilot

Through and Through Visual Studio App, Modern Installer, In-app Upgrades, Support64 bits and Copilot.

Bob Ward (Principal Architect, Microsoft), comes onstage.

Bob interrupts Shireesh and insists on a rematch! 

 Anna Hoffman (Principal GPM, Microsoft), comes onstage.

Bob utters “Anna, I’ve used SSMS from before you were born” and Anna warns Bob that she manages the team that Develops SSMS.

In Bob’s bravado, he is fine if she even competes against one of those SSMS dev team members, but has Bob met his match? Anna, calls Erin Stellato on stage and she retorts, “I’ve been using it as long as you, you know!”.

Erin Stelato (Principal Program Manager, Microsoft) enters the stage.

The competition begins and it is clear that Bob isn’t as familiar with the TSQL construct for the commands he’s attempting to execute on stage… perhaps the demo gods are frowning down on his over confidence. (or perhaps Bob is hamming it up because… well if you know Bob you will understand).

After Bob fumbles around looking for scripts and uses IntelliSense, he “lands the plane” but Anna and Erin seem un-impressed.

“That was sad… “ was Erin Stellato’s initial response.

Erin approaches the keyboard and proceeds to school Bob.

There was a mention of Git integration that slipped by too quickly to digest… Perhaps there will be more to see in this space soon…

Erin demonstrated the capacity to have your query tabs in SSMS in a location that suits you with many options to choose from (top, side, etc).

Vertical tabs is what Erin landed on (and that look really cool, honestly!)

Anna must have been up for the challenge today, or perhaps had a few extra quad shot espressos and utters, “you’re going down!” 

Erin next showcases the new Copilot capabilities by asking Copilot:

“What do I need to consider if I change the Compat Mode?” 

Copilot pulls together some of the common things to consider before making changes to compat level.

Erin points out that “I’m not going to change it in Production…” 

She’s going to “Save to Markdown file” so that she can have her team review the detail and talk about next steps 

Anna – “Let’s GO!”

In haste, Bob, comments that Copilot could possible know about Virtual Log Files (VLFs) and Erin is quick to make Bob eat his words. “It absolutely knows virtual log files…”

Erin’s prompt is “” and Copilot once again comes through and produces the output showing the number and size of the VLFs as included in the DMV.

But wait, there was a disruption that mimics an alien abduction… the lights flicker and sudden the projector resets and SSMS in now displaying Dark Mode!

I unison, Anna and Erin resound, “we have dark mode”

Shireesh Thota (Corporate Vice President, Azure Databases, Microsoft) takes the stage.

Shireesh once again decides on the winner and… Surprise, surprise, Bob once again came up short against his nimble opponents.

Bob gets a consolation prize, and the show goes on…

Every except for Shireesh Thota (Corporate Vice President, Azure Databases, Microsoft ) leaves the stage.

SQL Server enabled by Azure Arc

Shireesh shares details some of the things that Azure Arc can do for us. Including managing your SQL estate with improved observability, enhanced business continuity and governess. Also includes the ability to use defender and EntraID for on prem servers.

SQL Server Managed Instance improvements

Gives you:

 Build new AI dev/apps 

 Vector search capabilities in CosmosDB

Azure databases for building in the cloud

What are we doing in our database platforms and databases for building applications. 

13 million database in azure  (8:50 AM)

Azure SQL DB

Various Announcements

Hyper Scale Elastic Pools is also now Generally Available, providing for better performance at a lower cost as show by the sticker price in the following 2 slides:

Native vector (type and functions) Muazma & Davide Demo

Muazma and Davide were invited on staged to show how developers can use the new vector types and functions in to build intelligent applications.

Muazma Zahid (Principal PM Manager, Microsoft) and Davide Mauri (Principal Product Manager – Azure SQL DB) come out on stage.

The scenario that unfold is Davide is working with Contoso Insurance to build them a chatbot so that agents can more naturally explore the data about the customers through chatting with the bot to keep the customers happy and loyal. Davide emulate the agent asking about a customer he is about to meet with and the bot responds based on Davide’s access, of course protected with Row Level Security (RLS) adn Data Masking to show the agent only information about the customers that he is allowed to see and only those customers assigned to him.

The bot is answering by gathering data behind the scenes using SQL again the structured data. Davide asks the about about the fact that the solution is using Operational RAG where data is inserted into the SQL database as transactions as processed and are available to the chatbot to answer queries.

Information on a past complaint from the client was that they were upset about an unexpected increase in premium. This was accomplished with Similarity Search which is part of the new Vector Search capabilities. The dba developer a stored procedure that internally connects to openAI, transforms the query into an embedding, a vector, and then uses vector distance and new vector type to do similarity search.

Very impressive!

One last question to the chatbot opened the door to the next demo with Muazma and Aditya, which is, “Is there anything that can help to lower his claim.”

The chatbot responded indication that there is a “Safety Score” program to help lower their premiums based on their driving habits!

Aditya joins Muazma on stage and confirms that Hyperscale for Azure SQL Database has her back.

As customers that sign up for the Contoso Safety Score program, data from automobile IoT device will be streaming into the system in real time, and they project 1 trillion rows in 1.5 years. Aditya analyses the dataset and confirms that there are 1 trillion rows in the tracking table in the Contoso Safety Score database, and performance is not an issue. Hyperscale supports databases up the size of 128 TB.

Muazma also asks about multiple reads and writes and talks about Hyperscale being able to have 20+ replicas for reads and writes go to the primary replica so write performance isn’t affected by reads.

With that, Muazma calls Asad onto the stage.

Asad talks about Azure being the One Data Platform for you AI application

Asad make a few more Announcement on products becoming Generally Available or enter Preview, before introducing Shireesh back on stage to once again lead the conversation.

Shireesh talks about a customer story with Docusign using both Azure SQL Database Hyperscale and Cosmos DB

Mark Brown is then invited on stage to Demo Azure Cosmos DB Vector Search to Build a Semantic Cache for LLMs.

SEMANTIC CACHE 

Mark did a great job showing that by creating a semantic cache, requests that are semantically similar to previous searches can be answered without hitting the LLM, but instead are answered by the semantic cache, in turn saving time and money.

 GA – Dynamic Autoscal  in Cosmos DB has resulted in customers seeing a 15 –70% on cost savings 

Claire is invited on stage and she shows 2 performance benefits of building AI  applications on Azure PostgreSQL flexible server

Claire uses Azure local  AI and embedding runs locally within the Postgres Server. She then compare local AI with remote Azure Open AI service. She compares the performance of both of these side by side and the results show that the local ai embedding is 9 times faster than using the remote AI extensions.

Claire then shows how to improve search performance by usings Azure’s new DiskANN  Vector search over the regular HNSW (the open source, state of the art, vector search index) search. Unsurprisingly, the next generation of DiskANN being shown outperformed HNSW – it was 50 times faster!

Microsoft Fabric

Now this is a Microsoft Data Platform Community event… with less than 10 minutes left in this Microsoft led Keynote, and we finally hear about Fabric, and just one slide. It’s clear that they know their audience!

Demo – Bob & Connor Show: Query Performance using Fabric-mirrored SQL DB

After Bob and Connor adorn Shireesh with a Texas branded shirt, Bob gets right down to business with Connor asking him what he’s been up to, and Connor’s response is elegant… “I make SQL go fast “.

Connor talks on being focused on delivering the next version of Fabric

He talk on the fact that like SQL, Fabric uses a Front End node and requests go to back end nodes for scale and performance.

Connor went on to say that he’s be working on batch mode optimization in the following functions which result in 70% faster execution. And more batch mode optimization is coming! These Batch mode improvements help on all platforms and SQL Server vNext will include these optimizations.

Bob fired up SSMS and ran some queries leveraging some of these newly improved batch-mode optimized function to demonstrate that the queries we indeed running faster and the query plan’s clustered scan showed that they were indeed running in batch mode.

For this demonstration, Bob Mirror SQL from an OLTP system into compressed Delta Parquet files in Fabric and returned his results in 8 seconds – 1 TB of data comprising 196 million rows