Currently Cubulus scaling status is:
"scale up approach" = use multiple Python threads for running SQL queries. Works in theory: one single core CPU runs all the threads. Practical result: it does not crash under Apache Bench stress testing
-"scale out approach" = managed to set up a "cluster" of database processes running in same machine, but listening on different ports. It should be possible to run databases on different IP adresses, not just different ports. Fail-over and load balancing works when database processes are stopped & restarted. Again, it does not crash under load on single-core CPU
-"MySQL parallel query" - briefly tested on a friend's Intel Core Duo laptop. Setup : one Python thread, and MySQL 5.1 beta doing automatically parallel query on the partitioned table. The two cores were waiting for each other , most likely because partitions are not equal.
Cubulus code has to be modified and tested in order to scale properly.
Your hardware donations are welcomed! Ideally hardware would be a 4-way workstation, not necessarily new. Or a small server as long as the noise is compatible with family life.
Log in to post a comment.