UPDATE: The Denver Post is reporting that Denver Clerk and Recorder Wayne Vaden resigned today amid chaos in the Denver Election Commission which he oversees.
Of the hundreds of election problems that occurred across the country, few were more spectacular, at least visually, than the long lines of voters waiting to cast ballots in Denver. The mayor of Denver has called for an investigation and the auditor is calling to disband the election commission. The long lines were not the results of voting machine failure this time. Rather, the problems stemmed from the city’s decision to abandon precinct voting in favor of consolidated ‘vote centers’, which in turn led them to spend $85,000 on Sequoia Voting Systems electronic pollbooks.
Tens of thousands of voters ended up waiting in lines for hours because of the failure of their e-pollbook software. The mayor is livid and is demanding an investigation. Based on news reports, there were failures at every level: a failure of project management, of planning, of testing, of software development. It appears that Sequoia delivered software with serious design and implementation flaws and serious usability problems for poll workers that was inadequately tested before the election.
Denver’s decision to use vote centers, in which voters could vote at any center they chose on Election Day, required the city to use e-pollbooks to ensure that no one voted twice at two different locations. Each vote center had an e-pollbook, which was basically a laptop connected in real time to a central server over a T1 dedicated phone connection. The central server held a list of all eligible voters and whether they had voted yet or not. To sign in a voter, the poll worker entered their name into the e-pollbook, checked that the voter was eligible and hadn't voted yet, and then allowed them to vote.
Unfortunately, the server quickly became overloaded on Tuesday morning, and they weren't able to service voters quickly enough. Voting machines were sitting idle because of the bottleneck checking people in with the e-pollbooks.
At one point on Tuesday morning the administrators pulled the server out of service and replaced it with three servers, but that still wasn't enough to get the lines moving. Then they discovered the server was getting bogged down because poll workers were supposed to close their window after entering each voter name into the e-pollbook software and open a new window for the next voter, but some poll workers didn't close the window, and that left a session open on the server -- and after too many open sessions accumulated on the server, it became overloaded. Consequently, administrators were forced to take the server out of service and restart it more than once during the day, but the system was still too slow to keep up with the demand.
Moreover, the system was hard to use: poll workers who were comfortable with computers were able to process voters up to 8 times as fast as others who were not. Apparently computer-savvy workers were entering in only the person's first and last name into the search box (for instance), while the others were trying to fill out every single field listed, which was much slower.
The election commission's technology chief, Anthony Rainey, admitted that the software was never tested with 220 laptops trying to access 100,000 records in one day until the election, claiming there was no way to do that until you had 100,000 people show up. This is simply not true - of course you can test this stuff at scale, and it is irresponsible not to do so.
Sadly, these failures were predictable and avoidable. It's unfortunate that their technical people weren't paying attention to the USACM report on statewide voter registration databases, where they would have found discussion of many of these issues because maybe they would have made better decisions if they had been more aware of the technical issues.
Comment on This Article
You must login to leave comments...
Other Visitors Comments
You must login to see comments...