Communication analysis of the Avalanche Tirol App – Part 1

February 16, 2014

For some time now a mobile app for Andriod phones and iPhones is advertized which is called the official app of Tirol’s Avalanche Warning Service and Tiroler Tageszeitung (Tirol Daily Newspaper), so I installed it on my Android phone some days ago. Yesterday I went on a ski-tour (ski mountaineering) and on the way in the car I tried to update the daily avalanche report but it took really long and failed in the end. I thought that can’t be possible be, as the homepage of the Tyrol’s Avalanche Warning Service worked without any problems and was fast.

So when I was home again I  took a closer look the traffic the app sends and receives from the Internet … as I wanted to know why it was so slow.  I installed the app on my test mobile and traced the traffic it produced on my router while it launched the first time.  I was a little bit shocked when I look at the size of the trace – it was 18Mbyte big. Ok this makes it quite clear why it took so long on my mobile 😉  –> So part of the post series will be getting the size of the communication down , so I opened the trace in Wireshark and took at look at it. First I checked where the traffic was coming from.


So my focus was one the which was the IP address of and it is hosted by a German provider called Hetzner (you can rent “cheap” servers there).  As I opened the TCP stream I saw at once a misconfiguration. The client supports gzip but the server does not send gzipped.


Just for getting the value how much it would save without any other tuning I gzipped the  trace file and I got from 18.5Mbyte to 16.8Mbyte – 10% saved.  Than I extracted all downloaded files.  jpg files with 11Mbyte and png files with 4,3Mbyte … so it seems that saving there will help the most. Looking at the biggest pictures leaded to the realization that the jpg images where saved in the lowest compress mode. e.g. 2014-02-10_0730_schneeabs.jpg

  • 206462 Bytes: orginal image
  • 194822 Bytes: gimp with 90% quality  (10% saving)
  • 116875 Bytes: gimp with 70% quality (40% saving)

Some questions also arose:

  • Some information like the legend are always the same … why not download it only once and reuse until the legend gets update?
  • Some big parts of the pictures are only text, why not sent the text and let the app render it?
  • The other question is why are the jgep files 771 x 566 and the png files 410×238 showing the same map of Tirol? Downsizing would save 60% of  the Size (with the same compression level)
  • Why are some maps done in PNG anyway? e.g. 2014-02-10_0730_regionallevel_colour_pm.png has 134103 Bytes, saving it as jpeg in gimp with 90% quality leads to 75015 Bytes (45% saving)

So I tried to calculate the savings without minimizing the information that are transferred – just the representation and it leads to over 60% .. so instead of 18Mbyte we would only need to transfer 7Mbyte. If the default setting would be changed to 3 days instead of 7, it would go even further down, as I guess most people look only on the last 3, if even that. So it could come down to 3-4 Mbyte … that would be Ok, so please optimize your software!

I only wanted to make one post about this app, but then I found, while looking at the traffic, some security and privacy concerns I need to look into a bit closer …. so expect a part 2.


1 Comment »

RSS feed for comments on this post. TrackBack URI

  1. Hey Robert,
    as an app developer I pay attention to the network communications and the total data that has to be transferred, mostly because it can suck the battery quite fast. Definitely the analysis you made here was superb. This should help developers to also pay attention to optimisation details that are ignored by general rule.

    Keep it rocking!

    Comment by Antonio — November 23, 2015 #

Leave a comment

XHTML: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

Powered by WordPress
Entries and comments feeds. Valid XHTML and CSS. 75 queries. 0.239 seconds.