Atlanta MS BI Group September 28th Meeting

Come and join us on Monday, September 28th, for our next Atlanta MS BI Group meeting. The topic this time is “Enhancing Data Analysis and Predictive Analytics with NoSQL” delivered to you by RDA Consulting.

Presentation:

Enhancing Data Analysis and Predictive Analytics with NoSQL

  

Level: Intermediate

Date:

Monday, September 28th, 2015

Time

6:30 – 8:30 PM ET

Place:

South Terraces Building (Auditorium Room)

115 Perimeter Center Place

Atlanta, GA 30346

Overview:

The pervasiveness of Big Data has created significant challenges for companies that have historically utilized relational database management systems to perform data analysis and predictive analytics. This presentation outlines the challenges that stem from Big Data as it relates to deriving value from data, the types of NoSQL data stores available and how NoSQL can serve as a data platform that enhances the data analysis and predictive analytics capabilities of companies.

Speaker:

Cornell A. Emile is a Senior Software Engineer with RDA Corporation. Since 2001, he has designed, developed, and maintained a data-centric solutions within multiple industries such as Finance, Digital Marketing, State and Federal Government using Microsoft’s BI stack. His main interests include database performance tuning, ETLs, Business Intelligence and NoSQL.

Sponsor:

RDA provides specific enterprise solutions that integrate software, managed services, and the value of the cloud. We help you assess the best platforms, then design, build, implement and support tailored business solutions that optimize to your specific cost, scale, mobility and security needs.

Power BI Report Embedding on the Horizon

For years I’ve been harassing Microsoft to support embedding interactive reports connected to SSAS on custom apps. This feature has been sorely missing in the Microsoft BI stack. Yet, it’s a must-have feature that enables a wide range of integration scenarios, such as customer-facing reporting. Imagine your customers being able to ask natural questions for data exploration. This is a killer feature!

It looks like Power BI might finally bring some relief. I while back, I posted an Embedding Reporting feature wish on the Power BI discussion list. It now has 643 votes and it’s one of the most voted for features. Recently, when asked to provide feedback, I communicated to Microsoft how I envision report embedding to work:

  1. Developers must have the ability to embed reports/dashboards in applications. This should be conceptually similar to how they can do this now with ReportViewer. However, instead of an ASP.NET control, I’d imagine REST APIs with jQuery.
  2. The tricky part is security. It is paramount to be able to pass the user identity to on-premises SSAS models. Ideally, we don’t want to ask external users to register in Power BI. Instead, some kind of bulk licensing needs to be supported to allow requests to Power BI to go under a single trusted account but yet pass the user identity to SSAS.
  3. In a nutshell, the external user goes to the customer website and he’s authenticated using whatever authentication mechanism the app uses. Then, a call is made to Power BI to authenticate using a trusted account. When the user requests a report, the call must allow somehow passing the user identity. And, then this identity needs to flow to SSAS. Previously, to achieve this scenario and circumvent the SSAS limitation of supporting Windows security only, developers would add a CustomData connection string to the SSAS connection so that SSAS can use it for row-level security.

Recently, Lukasz Pawlowski (a Senior Product Manager formerly on the SSRS team and now heading the Power BI APIs) responded:

“Quick update here. While it’s not report embedding, we just announced a preview of an API to integrate dashboard tiles into applications. Read about it on the Power BI Developer blog”

So, it looks like the first wave of the embedding APIs has arrived. While it’s not what we need, I think we can expect full-featured embedding to arrive soon.

Azure D-Series VM Prices Go Down

I’m optimizing a SQL Server OLTP mirrored database hosted on an Azure VM. After load testing, it was clear that CPU is the primary resource constraint so I suggested an upgrade to a higher A-series machine alongside other reconfiguration and storage changes. Meanwhile, I’ve noticed that starting in October, the D-Series pricing goes down. Moreover, D-series include a SSD temporary drive to host the temp database and buffer extensions. And, according to Microsoft, D-series have “60% faster processors than the A-series”. In this case, to save the customer money, I recommended a D-Series VM with less cores than the A-series VM upgrade I had in mind.

Load Testing Azure Applications

Scenario: You’re conducting a capacity planning effort and load testing an Azure website (Azure App Service) that saves data to SQL Server running on Azure VM. You’ve created a load test, e.g. using Visual Studio, and you’ve realized that the throughput caps for no obvious reason. SQL Server and Azure performance counters indicate that both layers are not under pressure.

Solution: Although no specifics are provided by Microsoft, it appears that Azure throttles user loads to prevent denial-of-service attacks. Based on my observations, Azure detects that the requests come from the same IP and caps the requests. As a workaround, consider conducting the load test using Visual Studio Online. Visual Studio Online allows you to distribute the test across multiple agents that can use different IP addresses (configurable as a test property). In our case, this allowed us to conduct successfully a stepped load test until we reached a performance bottleneck. Interestingly, in this case the bottleneck we first reached was CPU utilization on the front end even although we scaled it to three large Azure web instances. SQL Server was churning along just fine with a single data disk. You results might differ of course.