Method and apparatus for decomposing I/O tasks in a raid system
The present invention implements an I/O task architecture in which an I/O task requested by the storage manager, for example a stripe write, is decomposed into a number of lower-level asynchronous I/O tasks that can be scheduled independently. Resources needed by these lower-level I/O tasks are dynamically assigned, on an as-needed basis, to balance the load and use resources efficiently, achieving higher scalability. A hierarchical order is assigned to the I/O tasks to ensure that there is a forward progression of the higher-level I/O task and to ensure that resources do not become deadlocked.
1. A method for decomposing a data access request from a file system into a plurality of lower-level I/O tasks in a RAID system, said method comprising:
representing logical combination and configuration of physical storage components of said RAID system as a hierarchical set of objects;
generating a parent I/O task from a first object in said hierarchical set of objects in response to said data access request from the file system;
dynamically allocating a resource for a child I/O task;
generating said child I/O task from a second object in said hierarchical set of objects to implement at least a portion of said parent I/O task, said child I/O task using the resources;
suspending the parent I/O task until said child I/O task is completed;
scheduling for execution said child I/O task by an I/O thread;
suspending the child I/O task by the I/O thread if the resource is not available;
executing said child I/O task in response to an occurrence of an event that the resource is available such that scheduling of said child I/O task is not conditional on execution of said parent I/O task; and
deallocating said resources allocated to said child I/O task when the child I/O task is completed;
resuming the parent I/O task upon an event indicating the completion of said child I/O task, said event indicating the completion occurred when a function vector is invoked by said child I/O task.
2. The method of claim 1 wherein the scheduling further comprises:
scheduling for execution said child I/O task in response to an occurrence of said event such that no synchronization points are required between said parent I/O task and said child I/O task.
3. The method of claim 1 further comprising:
queuing said parent I/O task and said child I/O task on separate I/O threads.
4. The method of claim 1 further comprising:
queuing said parent I/O task and said child I/O task on separate computing devices.
5. The method of claim 1 further comprising:
queuing said parent I/O task and said child I/O task to separate processes.
6. The method of claim 1 , further comprises:
scheduling said parent I/O task for execution in response to a second event; and
generating said child I/O task in response to execution of said parent I/O task.
7. The method of claim 1 further comprising:
queuing said parent and said child I/O tasks to an I/O thread, wherein said parent and child I/O tasks are independently schedulable entities.
8. The method of claim 1 wherein the step of suspending the child I/O task further comprises:
moving said child I/O task from an I/O thread to a wait queue.
9. The method of claim 8 wherein the step of suspending further comprises:
moving said child I/O task from an I/O thread to a wait queue.
10. The method of claim 1 wherein said child I/O task has no dependency on said first object, said first object being at a higher level in said hierarchical set than said second object.
11. The method of claim 1 further comprising:
generating, by said parent I/O task, an error-handling I/O task when said child I/O task encounters an error.
12. The method of claim 1 further comprising:
defining said parent and child I/O tasks having a standard interface.
13. The method of claim 1 further comprising:
defining said child I/O task including a start operation and a finish operation.
14. The method of claim 1 further comprising:
defining said parent I/O task including a start operation, a finish operation and a child-I/O-task-finished operation.
15. The method of claim 1 wherein said child I/O task is a bypass I/O task.
16. The method of claim 1 wherein said child I/O task is a first child I/O task, the method further comprising:
generating a second child I/O task to implement at least a portion of said parent I/O task; and
scheduling for execution said second child I/O task in response to an occurrence of a second event such that execution of said second child I/O task is not conditional on execution of said parent I/O task or said first child I/O task.
17. The method of claim 16 further comprising:
completing said parent I/O task upon completion of said first child I/O task and said second I/O task.
18. The method of claim 1 wherein said event is one of a disk I/O completion, a RAID group I/O completion, mirror I/O completion, plex I/O completion volume I/O completion or a request from a RAID client.
19. The method of claim 1 further comprising:
scheduling for execution a grandchild I/O task in response to an occurrence of a second event such that execution of said grandchild I/O task is not conditional on execution of said parent I/O task or said child I/O task; and
indicating to said child I/O task from said grandchild I/O task completion of execution of said grandchild I/O task.
20. The method of claim 1 wherein each I/O task is one of a disk read, a disk write, a volume write, a volume read, a mirror write, a mirror read, a plex write, a plex read, a group read, an array write, a parity reconstruction of a RAID group, a stripe parity reconstruction or a mirror verification.
21. The method of claim 1 , further comprising:
defining a plurality of states for each of the parent I/O task, the child I/O task, and a grandchild I/O task, and changing a state of the plurality of states from a first state to a second state, wherein said first and said second states are one of a perform I/O task state, a child I/O task pending state, an X 0 R task pending state, an I/O task finished state and an I/O task suspended state.
22. A storage system for decomposing a data access request into a plurality of lower-level I/O tasks in a RAID system, said storage system comprising:
a memory;
a network adapter;
an instantiator module stored in the memory configured to provide a hierarchical set of objects representing logical combination and configuration of physical storage components of said RAID system, to generate a parent I/O task from a first object in said hierarchical set of objects in response to said data access request received at the network adapter, to dynamically allocate a resource to a child I/O task, and to generate said child I/O task from a second object in said hierarchical set of objects to implement at least a portion of said parent I/O task, said child I/O task using the resources; and
an I/O manager stored in the memory configured to receive said parent and child I/O tasks and to schedule execution of said parent I/O tasks, to suspend the parent I/O task until said child I/O task is completed, to schedule for execution said child I/O task by an I/O thread, to suspend the child I/O task by the I/O thread if the resource is not available, to resume the parent I/O task upon an event indicating the completion of said child I/O task, said event indicating the completion occurring when a function vector is invoked by said child I/O task, to deallocate said resource allocated to said child I/O task when said child I/O task is completed, and to perform said data access request to accomplish an I/O data access to the RAID system of the storage system.
23. The storage system of claim 22 further comprising:
said I/O manager further configured to suspend processing of said child I/O task if said resource is not available.
24. The storage system of claim 22 further comprising:
said child I/O task configured to indicate its completion to said parent I/O task when said child I/O task is complete.
25. The storage system of claim 22 further comprising:
said instantiator module further configured to generate, from a call to said first object, an error-handling I/O task when said child I/O task encounters an error.
26. The storage system of claim 22 further comprising:
a first standard interface for said parent I/O task and a second standard interface for said child task.
27. The storage system of claim 22 further comprising:
a start operation and a finish operation for said child I/O task.
28. The storage system of claim 22 further comprising:
a start operation, a finish operation, and a child-I/O-task-finished operation for said parent I/O task.
29. The storage system of claim 22 further comprising:
said instantiator module further configured to instantiate a bypass I/O task.
30. The storage system of claim 22 wherein each I/O task is one of a disk read, a disk write, a volume write, a volume read, a mirror write, a minor read, a plex write, a plex read, a group read, an array write, a parity reconstruction of a RAID group, a stripe parity reconstruction and a minor verification.
31. A computer readable medium having stored thereon computer-readable instructions for execution on a processor to perform a method of decomposing a data access request into a plurality of lower-level I/O tasks in a RAID system, said method comprising the steps of:
representing logical combination and configuration of physical storage components of said RAID system as a hierarchical set of objects;
generating a parent I/O task from a first object in said hierarchical set of objects in response to said data access request from the file system;
dynamically allocating a resource for a child I/O task;
generating said child I/O task from a second object in said hierarchical set of objects to implement at least a portion of said parent I/O task, said child I/O task using the resources;
suspending the parent I/O task until said child I/O task is completed;
scheduling for execution said child I/O task by an I/O thread;
suspending the child I/O task by the I/O thread if the resource is not available;
executing said child I/O task in response to an occurrence of an event that the resource is available such that scheduling of said child I/O task is not conditional on execution of said parent I/O task; and
deallocating said resources allocated to said child I/O task when the child I/O task is completed;
resuming the parent I/O task upon an event indicating the completion of said child I/O task, said event indicating the completion occurred when a function vector is invoked by said child I/O task.