Zwiftalizer benchmarks frozen in time
In May 2019 I removed the log collection feature on zwiftalizer.com. I removed aging features that would have required a lot of maintenance to bring up to date - namely the benchmarks data collection. The main constraint is my limited free time, not hosting costs (but more on that later). My time is dedicated to my family, then career, then fitness. Hobby projects are frozen. It is unlikely that the benchmarks or the power sources reports will ever come back. They are just not priorities anymore. I will try to explain…
_More on the GDPR conversation- - I had the GDPR sit down with Zwift HQ to explain what happens to the log file data when someone puts it into Zwiftalizer. At the time (early 2018), the log file contained chat histories, your zwift ID, other users’ usernames, and Zwiftalizer collected the IP address of the browser used to ‘upload’ the log. This was a concern because of PII, GDPR and surveillance laws. I do not have the time to figure out the ins and outs of how this affects me as an individual, so I removed all code that was even remotely a concern. Now I do not even have to think about it. I can honestly say the site is 100% GDPR compliant because absolutely no information is collected by any back end whatsoever. There is no upload, no database and no storage, local or remote. Everything is client side in React and temporary. This is why the benchmarks stop at May 2019. For the tech nerds, the hosting is AWS CloudFront, S3 static hosting, Cert manger for SSL and Route53. (The services used to be API Gateway, Lambda, DynamoDB and streams to export the benchmarks to static json on S3 and CloudFront).
Why benchmarks got frozen - briefly, aggregating over nearly 3 years did not accurately account for performance optimizations made in the graphics engine. The long story - Averages of averages did reveal some interesting patterns, but honestly, who really cares? In 2015 I cared a lot, and so did the other folks getting set up because there was a lot of difference in AMD vs Nvidia vs Intel at that time. Now, pretty much anything made in the last 3 years will do a great job. AppleTV is popular, inexpensive and works great. iPads work great. Laptops work great. Gaming PC enthusiasts (self included) are a in a minority, probably. Or it is my belief that they will be soon. Finally, I felt PC enthusiasts skewed the benchmarks with their personal biases. I could not know if a log came from a crowded New York group ride, or a solo Richmond spin using camera 8. A controlled OpenGL benchmark like Heaven Unigine or Cinebench serves that audience.
The AWS bill - normally the hosting is less than a dollar a month. In April 2019 I got fancy and fired up an EMR data pipeline job and SageMaker to analyze my logs over time (manually uploaded). Of course, I went to bed that Friday night and forgot to shutdown the jobs and instances. When the bill came in May, AWS were very good about it and reimbursed me on condition that I stop and delete all instances, including dynamodb tables involved. This included the benchmarks. I kept a copy but have not had time to bring the database back up again. The current benchmarks report is made from a static data file.