Dernière mise à jour : 4 mai 2021

Power BI Summit is a multi-day event bringing speakers from the Microsoft Power BI team’s product group and community experts and MVPs from all around the world. This year’s virtual conference contained more than 100 sessions, covering all topics from Power Query, DAX, Dataflow, Visualization, Power BI Desktop, Power BI report server, Mobile experience, Embedded, Power BI Premium, Architecture, governance.

For those that did not get the chance to attend MDW, which is specialized in Power BI, have summarized a few interesting sessions for you.

Sessions covered in this article:

Govern your Power BI Environment with Azure Purview

Craig Bryden presented an overview of Azure Purview, the unified data governance service that helps you manage and govern your on-premises, multi-cloud and software-as-a-service (SaaS) data. He highlighted the importance of having a data catalog, as it allows:

  • Central repository for data assets

  • Data asset curation

  • Data asset discovery

  • Data lineage

  • Data quality

With this service, companies can:

  • Automate and manage metadata from hybrid sources.

  • Classify data using build-in and custom classifiers and Microsoft Information Protection sensitivity labels.

  • Label sensitive data consistently across SQL Server, Azure, Microsoft 365 and Power BI

  • Easily integrate all the data systems using Apache Atlas APIs.

Even though Power BI already has an integrated feature that allows you to visualize data lineage, Purview brings thigs to a next level:

  • Full end to end Lineage

  • Data asset discovery across workspaces

  • Asset curation beyond what is available in Power BI

Power BI Premium - Practical Tips for making the most of it

Power BI Dedicated Capacity requires some attention but doing so will result in the ability to do great things. During this session, Benni De Jagere touched base on some of the key activities’ organizations need to perform, to make sure their investment is one that pays off. He showed how using the Premium Capacity Metrics App as our base of operation, auditing, licensing, and some common dataset performance patterns, can make sure you're up to the task and using your premium capacity to its max!

Note that for this session Benni focused on Gen 1 (as it will still be for the next months, and most of it will be taken to gen 2).

What is Power BI Premium?

The Power BI Premium capacity is a dedicated capacity for your organization that provides a more complete feature set (some feature even release premium first, pro later (or not…). It allows for users to access artefacts without a Power BI subscription and it supports data residency by region (multi-geo) and customer-managed encryption keys for data at rest (BYOK). Note that it is not a shortcut to improve performance of your content in the workspaces.

Backend vs. Frontend v-cores:

  • Backend = everything that is related to refreshes, datasets and dataflows (regardless they are manual or triggered refreshes). Working with the actual data in the actual model

  • Frontend = resource that we use for rendering visuals, for getting query out of the data models, everything that is done for that interactive usage inside of power BI itself

RAM: working set of memory (e.g. P1 the highest point of memory we can use at one single point is 25 GB, for P2 50…). PBI Premium has the multimodal management (model management) where it balances the memory capacities itself, meaning that we can dump 50 GB of datasets into a P1 capacity and PBI will manage this for us.

Benni went through the Power BI Pro vs. Premium feature comparison list highlighting the features:

  • XMLA endpoint read/write connectivity

  • Application lifecycle management

  • Autoscale add-on availability (preview)

XMLA endpoint (the way for you to connect to the actual instance of AS that is working in the background and connect for example to Management Studio or any other tool that we may have). For this to work we need to have a READ WRITE enabled.

Tip: Enable Enhanced Dataflows Compute Engine (incredible difference in the performance when using dataflows and in the features that you can use). Make sure you limit the compute engine memory that this can use

You can check the complete list in the Microsoft official page, under the pricing section.

Here are the usable set of tips to assist daily tasks that he mentioned:

Check (and adjust) the capacity workload settings

  • Check the default capacity workload settingsà TIP: Do not overcommit the allocated resources

  • Questions to ask? Are all the workloads needed/used? Can we offload some workloads to other capacities?

How to set it up: Power BI Service > Settings > Admin portal

Question to ask when configuring the settings:

  • What is the current usage?

  • What is the potential growth that I may have?

  • How does this translates to the settings that I may need to have: Set the settings

Advice: Do not change the query timeout, max row counts… unless you have a very good reason to do so.

How do I know what is happening in my tenant? The only way right now is with the Premium Capacity Metrics Apps. Premium Capacity Metrics App v2 is your Ops base.

Good for Short term analysis

  • Cannot do real time connection

  • Not always as stable as we want to be

What to watch in your Capacity Metrics App

  • Memory evictions: eviction process: unload a dataset from a memory and load something else.

o There are perfectly normal, do not stress!

o Is there a threshold for when the evictions are too much? No, based on the models that you have, the usage that you have… but usually when the number goes to the high hundreds/thousands yes potentially you have an issue.

  • Hourly query distribution: What is happening with our queries? How long are they taking?

o We need to look at the distributions of our query.

o We want as max of the results to be either blue, black or orange, to see that we are not getting the long duration times.

  • Wait time distribution: How long does a query have to wait for other processes to finish before it gets its own going? What does the wait time distribution conclude?

o High wait times = busy reports on its own (prime candidates to be checked)

o What is the distribution for your wait time? As soon as they start diverging away from the blue, ALERT!

Benni is a firm believer in the value of Power BI Auditing. Here are the arguments he used to convince us of the power of this information.

  • Artefacts

  • Usage

  • User information

  • Licensing

Correlate these for insights into your capacity usage. Are users actively using their Pro Subscription? Are we having usage spikes at certain moments? Are we seeing behavior that need to be investigated?

Scheduled Refreshes

Check for congestions in the Capacity Refresh Summary [Admin portal > Capacity settings > Refresh summary]. Content Builders do not have access to this info. Extract refresh schedules and History and Build your report to provide them with this info so they know when they have a slot to set their own refresh.

  • Set up auditing to have a full overview of what is happening and correlate it to the usage

  • Activate large model support for model > 1GB

  • Offload non-production workloads to A Sku / PPU

Power BI Premium per User (PPU)

· Subscription model based on Premium Gen 2

· General availability on April 2021

· Aimed at smaller organizations, or offloading scenarios

Power BI Premium Gen 2