All releases tagged qosres
Release Notes: This release fixes various issues that were identified 1.0.0, and adds some extra capabilities to the core scheduling algorithm. Among others, the scheduler now implements spare capacity and/or dynamic reclamation policies through compile-time switches (see qres_config.h.in for details); the qres command-line utility has been enriched with an almost complete support to the operations available through the qreslib API; and the qres_setparams() function has been fixed, so that now the server period may be changed as well as the budget.
Release Notes: Assorted minor bugfixes.
Release Notes: A few configuration-related bugs have been fixed.
Release Notes: This release fixes a few configuration issues and minor bugs, and introduces the ability to reserve a certain amount of spare bandwidth in the supervisor configuration (useful for adaptive reservations).
Release Notes: This release introduces a couple of flags useful when creating servers. If QOS_F_PERSISTENT is enabled, a server is allowed to exist beyond detachment of the last task. If QOS_F_SOFT is enabled, a server's tasks are scheduled by the Linux default scheduler/policy when outside of the server reservation. Also, various stability issues in destroying servers have been fixed.
Release Notes: This releases fixes a few recursive spinlock bugs that are mostly harmless on UP, unless heavy spinlock debugging options or SMP are enabled in the kernel.
Release Notes: Issues related to the removal of the qresmod kernel module while servers were running have been solved. Also, a possible issue with versions of GCC lower than 4.1 has been avoided by providing a default configuration file that disables certain debugging capabilities making use of the __builtin_apply() and related GCC extensions.
Release Notes: The code has been completely refactored in order to be more maintainable and allow easier development of additional features (persistent servers, fast default server, soft reservations, etc.). The task dispatching mechanism has been redesigned and uses a new kernel patch. Instead of the fake task state that caused the scheduler to ignore the task, now the task is removed from the run queue when the server exhausts the budget.
Release Notes: This release of the Resource Reservation module comes with the support for multi-task servers enabled by default. This feature allows you to place a set of threads and tasks on a single server, thus allowing you to reserve a minimum guaranteed fraction of the CPU to them, as well as to restrict the maximum CPU fraction allowed to be subtracted from the system. Among others, the application rt-app.c is included , which highlights how real-time applications that would not be able to execute timely with the standard Linux kernel may execute correctly, taking advantage of AQuoSA.
