Statistics with 1200 athletes

15th Trofeo Judo Citta di Lavis#

On 20 and 21 January 2024 we held the 15th Trofeo in Lavis, with almost 1200 participating athletes: 500 on Saturday, 700 on Sunday. The event was entirely managed with Judo in Cloud, with exceptional results! You can see the information about the event at the following link.

As computer scientists, we were able to verify the ease of use, the autonomy and the reliability of our software, without any need for our intervention. Judo in Cloud passed with flying colours!

In the following paragraphs we show the usage data collected from the program, to prove that nowadays, even though sports halls are full of people and the interference is enormous, the cloud works when used by well-built software. The entire connection was based on a couple of SIM cards that provided internet inside a router. The connection was provided via wireless, with no cabling.

In short, Judo in Cloud handled a maximum of 4 requests per second with a maximum average latency of 0.5 seconds, with a maximum payload of 60kb incoming and 40kb outgoing (for our servers, so the opposite for the physical computers of the competition).

In simpler words, the connection at the sports hall was about 100 kb/s at the worst moments, and the program responded to every click in less than a second. At the same speed, a normal Facebook photo of 5 Mb would take almost a minute to load!

Requests#

We have two types of "computers" that we use to handle the incoming requests: "client" sends the web pages that you see on your PC or phone; "server" does all the calculations for the competition and connects to the database.

It is interesting to see how the requests (and also the other variables) follow a Gaussian distribution for each day.

Traffic#

Latency#

"Latency is the time interval between the moment an input/signal is sent to the system and the moment its output is available. In other words, latency is simply a measure of the response speed of a system". (source: Wikipedia).

In particular, here we show the mean values and the 50th and 95th percentiles. A 50th percentile means that at least half of the requests were resolved within that period of time at most.

It is interesting to note how during the night, when the servers were off and had to be started, the latency increases because it includes the startup time.