2018


Mongodb Query Planner Presentation
Some months ago I started working at MongoDB as Technical Services Engineer. It was a big and exciting change for me after several years in the relational world of MySQL. During that time a large percentage of cases from customers were related to bad query plans/wrong index selection. With MongoDB I found that the number of cases on that topic was very very (very) low. So I asked myself... why?

Prophet - Forecasting Our Metrics And Predicting The Future
Facebook recently released a forecasting tool called Prophet. Prophet can forecast a particular metric in which we have an interest. It works by fitting time-series data to get a prediction of how that metric will look in the future. In this blog post, we'll look at how Prophet can forecast metrics.

Services Monitoring with Probabilistic Fault Detection
Let's admit it, the task of monitoring services is one of the most difficult. It is time-consuming, error-prone and difficult to automate. The usual monitoring approach has been pretty straightforward in the last few years: setup a service like Nagios, or pay money to get a cloud-based monitoring tool. Then choose the metrics you are interested in and set the thresholds. This is a manual process that works when you have a small number of services and servers, and you know exactly how they behave and what you should monitor. These days, we have hundred of servers with thousands of services sending us millions of metrics. That is the first problem: the manual approach to configuration doesn't work.

2017


Batch Normalization for faster convergence
Batch Normalization is not a new technique recently discovered. It was described back in 1998 by Yan LeCun, et al in the whitepaper Efficient Backdrop. But I am the kind of person that prefers visual and straighforward examples that show how the theory applies in real life. It makes things easier to remember, don't you think?

Docker CLI Bash completion on MacOS
Docker CLI tool has many commands. If you are also managing a large number of containers, images and networks you will find yourself going back and forth to remember all names, identifiers and parameters. To get some help in this task we can use bash-autocompletion so it does all the magic for us when we push the "tab" key.

Automatic Cloudfront invalidation with Amazon Lambda
Serverless. That is the new magic word. Forget about containers, virtualization, operating systems, patching, scaling... Just write the code and upload it to Amazon Lambda.

Multipart uploads to S3
Sometimes I need to upload really large .zip files to S3. aws cli or the web interface works perfect for this task. The problem is, at least in my case, the bad quality of the internet connection. If for some reason it goes down I need to start the upload again from the very beginning.