Press "Enter" to skip to content

Author: Kevin Feasel

Power BI and Tabular Model Relationship Types

Published 2020-02-25 by Kevin Feasel

Marco Russo takes us through the different types of relationships we might encounter in Power BI and Analysis Services Tabular models:

A relationship can be strong or weak. In a strong relationship the engine knows that the one-side of the relationship contains unique values. If the engine cannot check that the one-side of the relationship contains unique values for the key, then the relationship is weak. A relationship can be weak either because the engine cannot ensure the uniqueness of the constraint, due to technical reasons we outline later, or because the developer defined it as such.

A weak relationship is not used as part of table expansion. Power BI has been allowing composite models since 2018; In a composite model, it is possible to create tables in a model containing data in both Import mode (a copy of data from the data source is preloaded and cached in memory using the VertiPaq engine) and in DirectQuery mode (the data source is only accessed at query time).

There is quite a bit of useful information in here.

Comments closed

Receiving Notifications when Azure Function Apps Fail

Published 2020-02-25 by Kevin Feasel

Gilbert Quevauvilliers shares how to receive notification e-mails when an Azure Function App fails:

Below are the steps to enable error notifications on Azure Function Apps

Follows on from my previous blog post How you can store All your Power BI Audit Logs easily and indefinitely in Azure, where every day it extracts the Audit logs into Azure Blob storage. One of the key things when working with any job that runs, is that I want to know when the job fails. If I do not have this and I assume that the data is always where, I could fall into a situation where there is missing data that I cannot get back.

Below explains how to create an alert with a notification email if an Azure Function App fails.

Read on for the step-by-step instructions.

Comments closed

Monitoring Availability Groups

Published 2020-02-25 by Kevin Feasel

Nisarg Upadhyay gives us some of the low-down on monitoring availability groups:

In my previous articles, I have explained the step-by-step process of deploying an AlwaysOn Availability group on SQL Server 2017. In this article, I am going to explain how to monitor AlwaysOn availability groups.

First, let’s review the configuration of the availability group we had deployed previously. To do that, open SQL Server Management Studio  Expand database engine from the object explorer  Expand “AlwaysOn High Availability”  Expand “Availability Groups.” You can see the availability group named SQLAAG. Under this availability group (SQLAAG), you can see the list of availability replicas, availability databases, and availability group listeners.

Click through for some tooling built into SQL Server Management Studio, as well as relevant Perfmon counters.

Comments closed

Understanding Azure SQL Database Elastic Jobs

Published 2020-02-25 by Kevin Feasel

Kate Smith takes us through some important concepts around Elastic Jobs in Azure SQL Database:

It is very important that the T-SQL scripts being executed by Elastic Jobs be idempotent.  This means that if they are run multiple times (by accident or intentionally) they won’t fail and won’t produce unintended results. If an elastic job has some side effects, and gets run more than once, it could fail or cause other unintended consequences (like consuming double the resources needed for a large statistics update).  One way to ensure idempotence is to make sure that you check if something already exists before trying to create it.

This takes some getting used to, but once you’re in the habit, you are much better off. Read on for more details on other key concepts.

Comments closed

Reading CPU Measures from the Ring Buffer

Published 2020-02-25 by Kevin Feasel

Taiob Ali explains what the CPU and memory measures are from the scheduler monitor ring buffer:

Here is a sample output of XML from sys.dm_os_ring_buffers where WHERE ring_buffer_type = N’RING_BUFFER_SCHEDULER_MONITOR’. What do those XML elements mean? In order to monitor CPU usages, you need to understand what each element means so you can use the values. I will explain each one in this blog post.

Read on for the list and what each means.

Comments closed

Entering Data into Power Query from Excel

Published 2020-02-25 by Kevin Feasel

Ed Hansberry shows a quick way to hand-enter some data into Power Query from Excel:

One of the cool things about Power BI is you have a nice “Enter Data” button on the home ribbon where you can quickly add data to your model. The table isn’t dynamic, but sometimes you just need a small table to finish your model to add descriptions or some other bit of data without creating another table in a source that needs to be connected to. Enter Data is perfect for this.

It did take a little bit of trickery to accomplish, but it’s pretty easy to do.

Comments closed

Dealing with Large SQL Scripts

Published 2020-02-24 by Kevin Feasel

Kevin Chant has some advice when you have to deal with a giant SQL script:

If you have been given a script that is thousands of lines long from a developer, the first thing I would ask is if they can split it up.

I say this because a lot of developers who write long scripts tend to have come from various backgrounds. Hence, some of them are used to developing on other programming languages.

So, they do not always appreciate that SQL is a set-based language. In addition, they do not always appreciate SQL Server is optimised for set based queries.

Sometimes you can break these scripts down, though there are of course good ways and bad ways to do so.

Comments closed

High-Throughput REST APIs with Dapper and Azure SQL DB

Published 2020-02-24 by Kevin Feasel

Davide Mauri builds out an example of a WebAPI project using Dapper to query Azure SQL Database:

I was able to execute 1100 Requests Per Seconds with a median response time of 20msec. If you can accept a bit higher latency, you can also reach 1500 RPS but the median response time becomes 40msec and the 95 percentile is set at 95msec. Database usage never goes above 20% in such cases…and in fact the bottleneck is the Web App (better, the Web App Plan) and more specifically the CPU. Time to scale up or out the Web App Plan.

By scaling up and out a bit, I was able to reach almost 10.000 request per second with just an HS_Gen5_4. Quite impressive.

I like Dapper as a micro-ORM. Products like it and FSharp.Data.SqlClient are good examples of how you can remove a lot of middleware goop without taking on the performance burdens of Entity Framework and Hibernate.

Comments closed