Making sense of API response times

Mathias Hansen, co-founder of Geocodio, on how they analyze their API response times:

For the longest time, we focused on average response times. It is a quick and simple metric that you can quickly break down by API endpoint and time window. […] While measuring average response time have their place they can also be deceptive.

The problem is that outliers go unnoticed that way, so better is to use percentiles:

Using percentiles, you will be able to say: β€œX% of API calls have a response time of Y milliseconds or less.”

A (rather big) query to quickly generate reports like these is included in the post πŸ™‚

Making sense of API response times →

Trending at Instagram

trending1trending2

We identified three main elements of a good trend:

  • Popularity – the trend should be of interest to many people in our community.
  • Novelty – the trend should be about something new. People were not posting about it before, or at least not with the same intensity.
  • Timeliness – the trend should surface on Instagram while the real event is taking place.

In this post we discuss the algorithms we use and the system we built to identify trends, rank them and show them in the app.

Trending at Instagram →