I was presented with the task of conglomerating the entirety of the inbound and outbound file transfer ecosystem of the area we were working in.

The challenge was complex in that all the connections were spread across different upstream and downstream networks, and across a variety of different servers, methods and interpretative codes.

I exclusively developed a daemon able to consume upstream feeds and feed downstream consumers seamlessly. The technology stack involved mainly Python, with a configurable set of rules and methods, and secure FTP, with a variety of SSH key combinations depending on the various connecting servers.

I had the task of solely and autonomously develop an initial multi-threading library, exploiting the Linux NUMA architecture.

The challenge was complex as no library of this kind existed in the area we were working in, so the technical solution had to be completely configurable in order to be connectable to existing APIs.

The initial version was based on various options to control the affinity of the threads, the sockets used and the cores available to each API and in each Linux server, which varied from PROD to the lower testing lanes, and was leveraged by several consumer applications.

The technology stack was composed of multi-threaded C++ (with STL and Boost) and the Linux NUMA library dynamically linked to the C++ code. The configuration was held on Sybase IQ databases and loaded once at runtime.