11 bookmarks. First posted by id1 16 days ago.
But there were two problems: The first was that the “CPU-intensive load” turned out not to be that CPU-intensive after all — a single task took a few milliseconds at worst. So most of the architecture was doing more harm than good. The second problem was that although it sounded like a highly scalable distributed system, it wasn’t one — it only ran on one machine. Why? Because all communication between asynchronous components was done using files on the local filesystem, which was now the bottleneck for any scaling. The original design didn’t say much about data at all, except to advocate local files in the name of “simplicity”. Most of the document was about all the extra architecture that was “obviously” needed to handle the “CPU-intensiveness” of the load.programming
7 days ago by ilprincipe