February 3rd, 2017

When it comes to high performance issue resolution, an out-of-the-box software solution stuck into a basic hosting platform just doesn’t cut it. For organizations prepping for Black Friday, or handling a massive onslaught due to unexpected media attention, or gearing up for end-of-year donations, performance, security and stability are beyond mission-critical: they’re existence-critical.

So when Election Protection, a U.S.-based, nonpartisan coalition founded to empower and enable voters, needed to boost performance and security of its issue tracking system for the 2014 midterm elections, it turned to GT.net. Our solution, detailed in the Online Issue Tracking Hosting case study, was built, tested and deployed to allow a network of thousands of volunteers the ability to enter voting data and issue tickets into the ‘Our Vote Live’ Request Tracker application. The data was then routed, classified and resolved as necessary by Election Protection staff.

The outcome? By any measure, a great success that underscored the importance of stress and load testing for high performance applications and websites. Thus when Election Protection needed similar support and capabilities for the 2016 U.S. presidential election, the group once again turned to GT.net.

Power to the People

Our solution was to turbo-charge what we had built in 2014, making sure that all voters in one of the closest and most fraught elections of the modern era would have their opinions heard. We were tasked with handling upwards of 100,000 tickets in 24 hours (or an estimated 120 tickets a minute at some points), 1,000 on-site end-users and 100 call-center end-users (routing issues from a toll-free hotline). Election Protection also required real-time analytics capabilities so that analysts could parse and examine trends in the data as it was collected.

Specifics included:

  • Seamless transitioning of the Election Protection application between the GT.net public cloud servers and a dedicated high performance cluster to ensure peak performance during high traffic periods
  • New workflows built for 2016 needs; 2014 workflows adapted as necessary
  • Stress-tests doubling the volume of our 2014 solution
  • Traffic simulation to troubleshoot any load or capacity issues
  • Improvements and customizations to base code to increase speed and functionality
  • Segmented access control by region
  • Expert consultation to determine and deliver to expectations for leadership, end-users and post-election analysts

Once again, GT.net delivered when it mattered.

1. Response times begin rapidly increase after reaching only 10% of throughput needed.
2. Request Tracker app crashes and reboots.

After recognizing that our initial load balancing solution wasn’t working to expectations (see spikes in graph above), GT experts iterated through refinements several times – improving code, tuning the server, adding additional indexes and re-syncing the database – until the system scaled to the necessary performance. The result was the steady and consistent flow of tickets apparent in the graph below.

1. Response times spike temporarily due to extremely large amount of concurrent requests in a short period of time.
2. Average response time is now within an acceptable range at the number of concurrent users needed.

Seeing our work put to the test – and perform at optimal levels – during the election was about as gratifying as one might expect. It also re-emphasized the importance of high-level stress and load testing to ensure reliability, availability and consistency throughout peak usage periods and beyond.

Organizations that need performance – but gamble on off-the-shelf systems and hosting – all too often see failures instead of results. If your organization needs to deliver, contact us to discuss our proven solutions, methodologies and capabilities. We’ll be happy to talk through any current or anticipated needs to ensure you’re at peak performance – always.