Storage transactions with predictable latency
The disclosure concerns at least one processor that can execute a polling group to poll for storage transactions associated with a first group of one or more particular queue identifiers. The disclosure concerns at least one processor is configured to: execute a second polling group on a second processor, wherein the second polling group is to poll for storage transactions for a second group of one or more particular queue identifiers that are different than the one or more particular queue identifiers of the first group, wherein the second group of one or more particular queue identifiers are associated with one or more queues that can be accessed using the second polling group and not the first polling group.
1. A computer-implemented method comprising:
a first polling group, executing on a first central processing unit (CPU) core, polling for identified storage transactions received in one or more packets in a first group of one or more queues, wherein the first group of one or more queues are associated with a first group of one or more particular queue identifiers and the first group of one or more queues are allocated in a memory of a network interface device;
a second polling group, executing on a second CPU core, polling for identified storage transactions received in one or more packets in a second group of one or more queues, wherein the second polling group polls for storage transactions for a second group of one or more particular queue identifiers and wherein the second group of one or more queues are allocated in the memory of the network interface device; and
performing the storage transactions identified by the first polling group and the second polling group, wherein
the first polling group performs a busy polling mode and defers interrupts to identify at least one input or output operation until after busy polling mode ends,
the busy polling mode comprises polling for the storage transactions in user space,
the first polling group polls the first group of one or more queues and not the second group of one or more queues,
the second polling group polls the second group of one or more queues and not the first group of one or more queues,
at least one transport protocol connection between a Non-volatile Memory Express over Fabrics (NVMe-oF) initiator and a NVMe-oF target is assigned to a queue of the first group of one or more queues, and
the storage transactions are based on NVMe-oF.
2. The method of claim 1 , wherein the first CPU core executes the first polling group and no other polling group.
3. The method of claim 1 , comprising allocating a transport protocol connection to a queue identifier and allocating the first polling group to the queue identifier of the transport protocol connection.
4. The method of claim 1 , comprising: based on termination of a transport protocol connection, de-allocating the terminated transport protocol connection from its queue identifier and de-allocating the first polling group from polling for input/output operations associated with the terminated transport protocol connection.
5. The method of claim 1 , wherein the storage transactions are received using a transport layer comprising one of: Transmission Control Protocol (TCP), InfiniBand, FibreChannel, or User Datagram Protocol (UDP).
6. The method of claim 1 , wherein the first polling group polls for storage transactions in user space.
7. The method of claim 1 , wherein a single queue of the one or more queues is allocated to multiple different transport protocol connections and the first polling group polls for storage transactions associated with the multiple different transport protocol connections.
8. A non-transitory computer-readable medium, comprising instructions that if executed by one or more processors, cause the one or more processors to:
execute a first polling group, on a first central processing unit (CPU) core, to poll for identified storage transactions received in one or more packets in a first group of one or more queues, wherein the first group of one or more queues are associated with a first group of one or more particular queue identifiers and wherein the first group of one or more queues are allocated in a memory of a network interface device;
execute a second polling group, on a second CPU core, to poll for identified storage transactions received in one or more packets in a second group of one or more queues, wherein the second polling group polls for storage transactions for a second group of one or more particular queue identifiers and wherein the second group of one or more queues are allocated in the memory of the network interface device; and
perform the storage transactions identified by the first polling group and the second polling group, wherein
the first polling group performs a busy polling mode and defers interrupts to identify at least one input or output operation until after busy polling mode ends,
the busy polling mode comprises polling for the storage transactions in user space,
the one or more queues are polled using the first polling group and not the second polling group, and
the storage transactions are based on Non-volatile Memory Express over Fabrics (NVMe-oF).
9. The non-transitory computer-readable medium of claim 8 , wherein the first CPU core executes the first polling group and no other polling group.
10. The non-transitory computer-readable medium of claim 8 , comprising instructions stored thereon, that if executed by one or more processors, cause the one or more processors to:
allocate a transport protocol connection between an NVMe-oF initiator and NVMe-oF target to a queue identifier and
allocate the first polling group to the queue identifier of the transport protocol connection.
11. The non-transitory computer-readable medium of claim 8 , comprising instructions stored thereon, that if executed by one or more processors, cause the one or more processors to:
based on termination of a transport protocol connection, de-allocate the terminated transport protocol connection from its queue identifier and
de-allocate the first polling group from polling for input/output operations associated with the terminated transport protocol connection.
12. The non-transitory computer-readable medium of claim 8 , wherein the one or more queues include at least one Application Device Queue (ADQ) accessible to the network interface device.
13. The non-transitory computer-readable medium of claim 8 , wherein the storage transactions are received using a transport layer comprising one of: Transmission Control Protocol (TCP), InfiniBand, FibreChannel, or User Datagram Protocol (UDP).
14. An apparatus comprising:
an interface
at least one central processing unit (CPU) processor coupled to the interface, the at least one CPU processor configured to:
execute a first polling group, on a first core of the CPU, to poll to identify storage transactions received in one or more packets in a first group of one or more queues, wherein the first group of one or more queues are associated with a first group of one or more particular queue identifiers and the first group of one or more queues are allocated in a memory of a network interface device;
execute a second polling group, on a second core of the CPU, to poll to identify storage transactions, wherein the second polling group is to poll for storage transactions for a second group of one or more particular queue identifiers and wherein the second group of one or more queues are allocated in the memory of the network interface device; and
process the storage transactions identified by the first polling group and the second polling group, wherein
the first polling group is to perform a busy polling mode and defer interrupts to identify at least one input or output operation until after busy polling mode ends,
the busy polling mode comprises polling for the storage transactions in user space,
the one or more queues are polled using the first polling group and not the second polling group and
the storage transactions are based on Non-volatile Memory Express over Fabrics (NVMe-oF).
15. The apparatus of claim 14 , wherein the first polling group is to execute on the first core of the CPU and the first core is to executes no other polling group.