Why we dropped Microsoft Power BI and embraced AWS QuickSight

Views are my own.

Microsoft Power BI is a great tool for end users because it really captures a users attention with beautiful visualizations and has a very nice (probably the best) native iOS and Android App on the Business Intelligence arena.

I tried a lot a B.I. tools out there, ranging from IBM Cognos, Jaspersoft AWS edition, Microsoft Power BI, Google Data Studio and more recently AWS QuickSight.

At the beginning of 2016 we where looking for a cloud-first Business Intelligence solution for small companies that just can’t afford a legacy licensing + on-premises product. That’s where we happen to cross with Power BI and immediately tried it out and in very few days achieved what we wanted: bar charts and tables with information pulled out from MySQL.

So we created some demos and captured the attention of some companies that, after testing side-by-side Power BI and Google Data Studio, decided for the latter.

One company choose Power BI for having an iOS App. So we built that project using AWS MySQL Aurora and made most of the calculations on it. Our first obstacle with Power BI is that it is designed for Windows developers because the only tool to build the reports and dashboards runs on Windows.

Power BI Desktop installation was not straight forward (a full set of prerequisites if you are not running on Windows 10). Then it comes the updates and patches (typical Microsoft style), and the connectors because Power BI Desktop does not connect natively to MySQL, it relies on MySQL Connector for .NET.

While on developing stage, we manually published the reports and dashboards from Power BI Desktop to Power BI Web.

Until today, if you are using MySQL, the only way a data set can be scheduled refreshed runs on Power BI Gateway, a program that relies on Windows! So we tried to install it on Windows Server 2008 … and we couldn’t, lot of errors appeared. Then we tried installing it on Windows 10 (a user PC), and it was not straightforward, several errors again appeared and we found latest version of the Gateway buggy so we were lucky to find an older version.

For three days the gateway worked fine, then suddenly stopped working and after searching on the community blogs we found a lot of other people had the same trouble.

Schedule refresh history using Power BI Gateway with MySQL.

The gateway operation has been cancelled explicitly by the user or from a timeout.

… credentials were deleted every time we had the above error.

Cannot connect to the mashup data source.

Power BI Gateway October update.

Another issue came up: dashboards were not displaying the same information as reports … what??

Lets go to Power BI basics: To get a dashboard in Power BI you first need to build a report and then “pin” it as a dashboard. This two pieces of Power BI are treated as independent components, and once you change something in a report those changes are not applied to the dashboard… so they are not in sync automatically… ¿are you kidding me?

I tried tweeting and contacting Microsoft but the responses just left me with the same taste when my Windows XP crashed with a blue screen … your are there alone and the frustration is just unbearable.

Finally with Power BI, there was another thing we couldn’t really address at that time but for sure it will morph into a growing snow ball on the future: our MySQL instance, (we use the professional AWS version called Aurora with a t2.small instance type), was crashing daily when Power BI Gateway was attempting to run the scheduled refresh. We even got an automated email from AWS letting us know of this issue.

AWS automated email, Aurora instance — Out of Memory

Months before this particular project we experienced the same issue on a bigger instance, but did not pay much attention:

10 reasons we had to move to AWS QuickSight

AWS QuickSight

When my frustration with Power BI reached its highest levels, I decided to fill the contact information page with AWS QuickSight. They replied immediately and in less than a week I was attending a one-on-one presentation.

I expressed them my concerns on QuickSight:

1. The iOS QuickSight App was only available in the US, and we need it for Mexico. Also last update to the iOS App was on January, nine months ago, so I was not sure of how serious this app was for AWS and QuickSight roadmap.
2. We needed a “developer” account free of charge to build the reports/dashboards.
3. We wanted to work with the smallest, currently t2.small, Aurora instance and be sure nothing crashes when the schedule refreshes the data sets.

After attending the presentation, I decided to give QuickSight a try, further more, it was clear that our account was not being charged for this because QuickSight has 1 free user for unlimited time.

So here is a bit of what we achieved:

1. Data sets are really straightforward and SPICE (the in-memory calculation engine) is activated with just a click, no setups.
2. Data visualizations (called visual types on QuickSight) are limited but offered what we needed. Currently there are no maps, plain tables and other types of visuals where Power BI has a superb collection.
3. Making a dashboard from an analysis is easy.
4. Setting up a schedule refresh for data sets its easy and works every time.
5. Data set refresh is optimized to make the lower impact on your database workload.

Best of all, we where able to move all the project from Power BI to QuickSight in about two weeks, almost with no troubles on the way.

After a week working with QuickSight, I decided to email Jeff Barr on a Sunday afternoon, and surprisingly I got a reply that same day.

I wrote him about my thoughts on QuickSight and he kindly contacted me with people working directly on QuickSight at Seattle.

The following Monday I was surprised and the iOS QuickSight App was now available on the Mexican App Store.

That same week, I was in the middle of an hour long call with an AWS QuickSight manager, screen sharing what we thought was missing on QuickSight.

Choosing the right horses to ride

As Steve Jobs said: You have to chose the right horses to ride:

So we are choosing AWS QuickSight as a technical vector that has a future, which we think is on its spring.

Disclosure: Jaime Navarrete works as a Solutions Architect at Factor BI, a startup working on database synchronization, Business Intelligence and tailored made enterprise web applications. Views are my own.