AWS QuickSight Test Drive

While there’s no shortage of data visualization tools, there is one commonality among the more prominent solutions: significant cost. This, coupled with the fact more and more companies are increasing their adoption of Amazon data solutions, I figured the time had come to kick the tires on AWS’s QuickSight data visualization tool. Here is a quick synopsis of what I discovered.

 

The Good

As advertised, QuickSight was really easy to setup and lends immediate dataviz value for users of Amazon data products. Here are some aspects of QuickSight that worked quite well; spanning the areas of data connectivity, speed , data blending, user interface, and cost.

 

Data Connectivity

It should come as no surprise that Amazon’s data visualization tool plays nice with other Amazon solutions. This includes:

  • S3 – Amazon’s Simple Storage Service, essentially a massive-scale object store.
  • Athena – Amazon-hosted Presto, which provides a nifty SQL interface to S3. This essentially turns S3 into a very simple (and inexpensive) data lake that is easily accessible to “traditional” BI/data visualization tooling.
  • RDS, Redshift, Aurora – Transactional, data warehouse, and high-speed MySQL databases.

QuickSight can also connect to non-AWS data stores, including files and standard databases such as MySQL, MariaDB, PostgreSQL and Microsoft SQL Server.

 

Speed and Scale

Like many other data visualization tools, QuickSight stores content within memory to lend additional performance pep. Dubbed “SPICE” (Super-fast, Parallel, In-Memory Calculation Engine) this is the engine that essentially imports source information into a memory-resident cube.

 

Blended Data

It took some poking around to figure out, but QuickSight can in fact provide blended data reports. The taxonomy of QuickSight objects follows:

quicksight_objects

A dashboard can have an analysis grouping of one or more visual elements; each of which is interactive. When it comes to visualizations, QuickSight ships with fairly standard visualization types one would expect to encounter in a dataviz tool. Usual suspects include: bar charts, lines charts, area charts, pivot tables, scatter plot, treemap and heatmap:

dashboard

 

Web-Based and Mobile-Friendly

Certain traditional data visualization solutions require a desktop application to “author” new dashboards. (Tableau, Qlik, and Microstrategy to name a few.) QuickSight is a web pure play; so no desktop software is needed. Additionally, QuickSight has a responsive HTML5 interface which adapts itself to desktop, smartphone or tablet user agents. Unfortunately, creating or even tweaking existing dashboards doesn’t appear to be an option for mobile warriors at this time.

 

Cost

Amazon markets QuickSight as a BI solution that costs 1/10th that of traditional solutions. How does this stack up against something like Tableau Online? Assume we’ve got ten users, two of which need to author dashboards, this is how things stack up:

TableauVSQS

Perhaps not an true apples-to-apples comparison, but the table does convey the ballpark price ranges for a small team of ten users who need to author and consume data visualizations from the cloud.

 

The Bad

QuickSight may have been quick and painless to set up, but there’s definitely some lacking functionality. I’ll spare the suspense and jump right into it:

  • No JSON Format Support – QuickSight supports tab and comma-seperate files, native Microsoft Excel files, and common logging formats. It doesn’t however support JavaScript Object Notation, which is the primary data format for modern web service APIs.
  • No Report Scheduler – If you want to email daily or weekly reports to information consumers, you’re going to have to wait. The only way to consume dashboard data is to interactively login to the web user interface. (You can, however, refresh SPICE hypercubes on a schedule; at least for some data sources)
  • No persistent data value labels – Perhaps this was error on my part, but I was unable to turn on data labels. For instance, if I wanted to view the actual values for the slices of a pie chart or a bar graph, I had hover the mouse over every segment, one-by-one.
  • No Embedding of Dashboards – Syndicating a dashboard to another information delivery portal (e.g. Sharepoint, Confluence, Slack, etc) is a common enterprise request. Unfortunately it’s a feature we’ll have to wait for as this is not yet available in the current release.
  • No Geo Capabilities – Tabled are plans for geographic visualization as there is no geo/map/coordinate visualization functionality at this time.
  • No VPN Connectivity to the On-Prem World – I assumed an AWS solution would enable customers to build virtual private networks that bridge the customer’s VPC into internal corporate networks, but no such luck. Therefore if you want QuickSight to report on your on-prem databases, you’ll need to expose them to the public Internet; an ask that will undoubtedly make your information security team cringe.

 

The Ugly

For the record, “ugly” refers to all things that land somewhere between good and bad. It’s basically the list of functionality that could really be better. Without further adieu, the list includes:

  • It’s on the Roadmap! – Many of the missing features above are planned to be released in future editions as explained by Amazon support representatives patrolling the QuickSight forums. Unfortunately, Amazon isn’t necessarily forthcoming with dates. Some of the more popular feature requests are said to be released in “the coming months.” I suppose I’ll stay tuned for the next 30 to 364 days…
  • Limited SaaS Connectivity – The only SaaS solution QuickSight connects to at this time is Salesforce.com. While one could argue that SaaS connectivity isn’t a data visualization tool’s job, Tableau or Domo would certainly beg to differ.
  • Misc Nitpicks – Salesforce objects must be selected one-at-a-time, so a SOQL-based query would be much more efficient. The aforementioned QuickSight Forums are, for some reason, disparate from the “main” Amazon support forums. Why is beyond me.

 

Bottom Line

If you primarily need to perform rudimentary visualization of potentially large-scale information housed within Amazon data solutions, then QuickSight may be a solid contender. However, if you have “traditional” enterprise requirements such as the need to directly connect to SaaS solutions and web services, or if you need direct connectivity to on-premises solutions such as Oracle databases or legacy CRM solutions, you may need to keep shopping around. At least for now.

 

Leave a Reply