Predictable client latency via background file system operations

A non-transitory processor-readable storage medium comprising one or more instructions which, when executed by at least one processing circuit, causes the at least one processing circuit to manage an amount of storage written from a transaction group to a set of data storage drives, by:

• a) creating a plurality of transaction classes of operations within each of a plurality of write transaction groups representing a storage pool allocator (SPA) synchronization cycle by which in-memory data is written persistently to one or more data storage drives of the set, wherein each transaction class is associated with how that respective transaction class is handled by a file system;
• b) discovering an initial drive capability for the data storage drives based on instruction operations per second (IOPS) and a data rate per second (MBPS) per data storage drive and assignable to each of the transaction classes when a storage pool is allocated via a SPA synchronization;
• c) assigning underlying drive capability to each transaction class based on a corresponding transaction class capability weight relative to the capability weights for each other transaction class;
• d) adjusting each transaction class capability weight based on the IOPS, the MBPS, and a time slice at every SPA synchronization;
• and
• e) lending an unspent time slice from a transaction class whose transaction class capability weight was adjusted to a plurality of remaining transaction classes whose transaction class capability weights are yet to be adjusted, via a staged processing distribution during a SPA synchronization of the SPA.

A system for managing an amount of storage written from a transaction group to a set of data storage drives, the system comprising:

• a) a module configured to create a plurality of transaction classes of operations within each of a plurality of write transaction groups representing a storage pool allocator (SPA) synchronization cycle by which in-memory data is written persistently to one or more data storage drives of the set, wherein each transaction class is associated with how that respective transaction class is handled by a file system, and configured to discover an initial drive capability for the data storage drives based on instruction operations per second (IOPS) and a data rate per second (MBPS) per data storage drive and assignable to each of the transaction classes when a storage pool is allocated via a SPA synchronization;
• b) a module configured to assign underlying drive capability to each transaction class based on a corresponding transaction class capability weight relative to the capability weights for each other transaction class;
• c) a module configured to adjust each transaction class capability weight based on the IOPS, the MBPS, and a time slice at every SPA synchronization;
• and
• d) a module configured to lend an unspent time slice from a transaction class whose transaction class capability weight was adjusted to a plurality of remaining transaction classes whose transaction class capability weights are yet to be adjusted, via a staged processing distribution during a SPA synchronization.