Dynamic weight accumulation for fair allocation of resources in a scheduler hierarchy
Methods and systems are presented for allocating resources based on dynamic weight accumulation performed in a bottom-up fashion in a scheduler hierarchy of a storage system. One method includes assigning weights to leaf schedulers at a bottom level of schedulers in a scheduler hierarchy comprising a plurality of levels of schedulers. Between two levels, each parent scheduler is associated with a unique plurality of children schedulers. For each leaf scheduler that is active, the method includes propagating a corresponding weight of a corresponding leaf scheduler to every higher level in the scheduler hierarchy, such that a corresponding scheduler at a corresponding level is associated with an accumulation of weights of its descendent schedulers. The method includes distributing resources assigned to the scheduler hierarchy based on accumulated weights at each level, such that the corresponding scheduler is proportioned resources based on the accumulation of weights of its descendent schedulers.
1. A method for resource allocation, the method comprising:
assigning a plurality of weights to a plurality of child schedulers, each child scheduler associated with multiplier the child schedulers being in a scheduler hierarchy with a plurality of parent schedulers, wherein each of the plurality of parent schedulers is associated with a unique group of the child schedulers;
for each child scheduler that is active, propagating a value based on the assigned weight of the child scheduler upwards in said scheduler hierarchy through a respective chain of schedulers to cause a given scheduler at a given level in the respective chain of schedulers to be associated with an accumulation of values based on the weights of descendent schedulers of the given scheduler along the respective chain;
for the given scheduler at the given level, factoring in the multiplier applied to said accumulated values of the descendant schedulers in the respective chain of schedulers to generate a multiplied value;
propagating said multiplied value upwards through said respective chain of schedulers; and
distribute a given set of resources assigned to said scheduler hierarchy based on multiplied value at each level of schedulers to cause the schedulers in the scheduler hierarchy to be proportioned resources from said given set of resources based on said multiplied value.
2. The method of claim 1 , further comprising applying the multiplier to the given scheduler, wherein the given schedule is located above a level containing the child scheduler in the scheduler hierarchy.
3. The method of claim 1 , wherein said multiplier comprises a number greater than or equal to 1.
4. The method of claim 1 , wherein said scheduler hierarchy comprises:
a top level comprising a foreground input/output (IO) scheduler;
a first sublevel below and adjacent to said top level, wherein said first sublevel comprises a Read Admit Scheduler, a Write Admit Scheduler, and a Continuing Scheduler;
a second sublevel below and adjacent to said first sublevel, wherein said second sublevel comprises a plurality of folders, each of the plurality of folders being associated with one of said Read Admit Scheduler and Write Admit Scheduler; and
a third sublevel below and adjacent to said second sublevel, wherein said third sublevel comprises a plurality of volumes, each of the plurality of volumes being associated with one of said plurality of folders and said Continuing scheduler.
5. The method of claim 4 , wherein said first sublevel further comprises a Remote Write Admit Scheduler.
6. The method of claim 1 , further comprising:
triggering accumulation of said plurality of values periodically based on a predetermined period.
7. The method of claim 6 , wherein said predetermined period comprises 200 ms.
8. The method of claim 1 , further comprising:
triggering accumulation of said plurality of values after a change in status of a scheduler in said scheduler hierarchy.
9. The method of claim 1 , wherein said given set of resources comprises a plurality of CPU cycles.
10. The method of claim 1 , wherein said scheduler hierarchy comprises:
a top level comprising a data disk input/output (IO) scheduler.
11. The method of claim 1 , wherein, for each child scheduler that is active, the value based on the assigned weight is equivalent to at least one of the assigned weight or a dynamic weight based on the assigned weight.
12. A storage system comprising:
a non-volatile memory (NVRAM) for storing incoming write requests;
a solid state device (SSD) configured as read cache memory;
a hard disk drive (HDD) for permanent data storage; and
a central processing unit (CPU), wherein a CPU scheduler of said CPU is to:
assigning a plurality of weights to a plurality of child schedulers, each child scheduler associated with multiplier, the bottom schedulers being in a scheduler hierarchy with a plurality of parent schedulers, wherein each of the plurality of parent schedulers is associated with a unique group of the bottom schedulers;
for each bottom scheduler that is active, propagate a value based on the assigned weight of the bottom scheduler upwards in said scheduler hierarchy through a respective chain of schedulers to cause a given scheduler at a given level in the respective chain of schedulers to be associated with an accumulation of values based on the weights of descendent schedulers of the given scheduler along the respective chain; and
for the given scheduler at the given level, factoring in the multiplier applied to said accumulated values of the descendant schedulers in the respective chain of schedulers to generate a multiplied value;
propagate said multiplied value upwards through said respective chain of schedulers; and
distribute a given set of resources assigned to said scheduler hierarchy based on multiplied value at each level of schedulers to cause the schedulers in the scheduler hierarchy to be proportioned resources from said given set of resources based on said multiplied value.
13. The storage system of claim 11 , wherein said scheduler hierarchy comprises:
a top level comprising a foreground input/output (IO) scheduler;
a first sublevel below and adjacent to said top level, wherein said first sublevel comprises a Read Admit Scheduler, a Write Admit Scheduler, and a Continuing Scheduler;
a second sublevel below and adjacent to said first sublevel, wherein said second sublevel comprises a plurality of folders, each of the plurality of folders being associated with one of said Read Admit Scheduler and Write Admit Scheduler; and
a third sublevel below and adjacent to said second sublevel, wherein said third sublevel comprises a plurality of volumes, each of the plurality of volumes being associated with one of said plurality of folders and said Continuing scheduler.
14. The storage system of claim 11 , wherein said CPU scheduler accumulates said plurality of values periodically based on a predetermined period.
15. The storage system of claim 11 , wherein said CPU scheduler accumulates said plurality of values based on an event.
16. The storage system of claim 12 , wherein, for each bottom scheduler that is active, the value based on the assigned weight is equivalent to at least one of the assigned weight or a dynamic weight based on the assigned weight.
17. A non-transitory computer-readable storage medium storing a computer program for allocating cycles of a CPU (central processing unit), wherein the computer program, when executed by the CPU is to cause the CPU to:
assigning a plurality of weights to a plurality of child schedulers, each child scheduler associated with multiplier, the bottom schedulers being in a scheduler hierarchy with a plurality of parent schedulers, wherein each of the plurality of parent schedulers is associated with a unique group of the bottom schedulers;
for each bottom scheduler that is active, propagate a value based on the assigned weight of the bottom scheduler upwards in said scheduler hierarchy through a respective chain of schedulers to cause a given scheduler at a given level in the respective chain of schedulers to be associated with an accumulation of values based on the weights of descendent schedulers of the given scheduler along the respective chain;
for the given scheduler at the given level, factor in the multiplier applied to said accumulated values of the descendant schedulers in the respective chain of schedulers to generate a multiplied value;
propagate said multiplied value upwards through said respective chain of schedulers; and
distribute a given set of resources assigned to said scheduler hierarchy based on multiplied value at each level of schedulers to cause the schedulers in the scheduler hierarchy to be proportioned resources from said given set of resources based on said multiplied value.
18. The storage medium of claim 17 , wherein said scheduler hierarchy comprises:
a top level comprising a foreground input/output (IO) scheduler;
a first sublevel below and adjacent to said top level, wherein said first sublevel comprises a Read Admit Scheduler, a Write Admit Scheduler, and a Continuing Scheduler;
a second sublevel below and adjacent to said first sublevel, wherein said second sublevel comprises a plurality of folders, each of the plurality of folders being associated with one of said Read Admit Scheduler and Write Admit Scheduler; and
a third sublevel below and adjacent to said second sublevel, wherein said third sublevel comprises a plurality of volumes, each of the plurality of volumes being associated with one of said plurality of folders and said Continuing scheduler.
19. The storage medium of claim 17 , further comprising wherein the computer program is further to cause the CPU to:
trigger accumulation of said plurality of values periodically based on a predetermined period.
20. The storage medium of claim 17 , wherein, for each bottom scheduler that is active, the value based on the assigned weight is equivalent to at least one of the assigned weight or a dynamic weight based on the assigned weight.