It’s the race that stops Australia. TAB claims they handle 3000 transactions per second at the peak of the betting splurge – across desktop and mobiles in all parts of the country. You can imagine, preparing for peak digital performance is a major priority for all betting agencies at this time of year.
As usual, my interest in how digital businesses handle themselves during big events led me to set up some performance testing.
Specifically, I looked at:
- The top 10 betting agencies in Australia
- Homepages of each agency on desktop and on mobile (iPhone 6)
- Response times for those homepages
- Number of hosts
- Number of objects
- Site weight (page size)
Encouraging a little old-school state rivalry, I also ran the tests from two different locations to see whether Melbourne or Sydney punters enjoyed a better punting experience.
1) Consistency in performance
Response times fluctuated slightly more than I expected. Averaging about 9 seconds for the most part, the peak response time was just prior to the race (Nov 1 @14:00) – 14 seconds).
This is likely to be when systems are at the biggest amount of load.
2) CDN performance issues
The majority of hosts performed well. Except one. And it didn’t just perform badly it basically imploded.
Who was it? I’m not going to disclose, but it did relate to a lesser known CDN provider (Not one of the enterprise-grade CDN providers with which we partner).
Basic learning. You get what you pay for. If you want to survive peak performance don’t skimp on CDN performance!
What this meant for the site in question was extreme response time increases which would have rendered the site unusable for many.
3) Melbourne vs Sydney user experience
Melbourne won the desktop response time battle, but Sydney took the crown for mobile experiences. As a former Melbournian, calling a dead heat again is painful. But what it does show is, you cannot focus on desktop, or mobile alone. You have to look into geographical differences to get a better picture.
For those in the digital performance game themselves, you might be interested in reading a bit more about how the agency sites compared/performed and how they could have improved:
Mobile was faster than desktop by nearly 2 seconds (8.3 Mobile vs 10.7 desktop)
- The fastest response time was 3.18 secs and the slowest 18 seconds.
- Mobile was faster despite 4G being a slower connection than the high-availability data centre connection.
- Testing using Dynatrace last mile (real pc network) would probably show a much higher advantage to mobile
Number of hosts were almost the same. (28 mobile vs 29 desktop)
- Initially I was surprised to see 28 external hosts for a peak event like the Melbourne cup. Apple has only 3 external hosts and typically serves a site in 1 second.
- Amazon, also known for fast response times, had 22 hosts for their prime day sale, so the number of hosts is comparable.
Mobile was heavier despite less objects (1.6MB vs 1.4MB)
- This is surprising. You’d expect a site that has less objects would be lighter. You’d also expect performance teams to focus on putting their sites on diets to make sure they are ready to race.
I hope those in Australia enjoyed the day, and you managed to win a few dollars like my mum did last year — $22,000 — after reading my blog on who to bet with! Here’s the story and the blog!
For more on this topic, you may enjoy this piece — How Amazon Preps it Retail Site Performance for Prime Day.