Optimization for real-time, parallel execution of models for extracting high-value information from data streams
A computer system identifies high-value information in data streams. The computer system receives a filter graph definition. The filter graph definition includes a plurality of filter nodes, each filter node including one or more filters that accept or reject packets. Each respective filter is categorized by a number of operations, and the one or more filters are arranged in a general graph. The computer system performs one or more optimization operations, including: determining if a closed circuit exists within the graph, and when the closed circuit exists within the graph, removing the closed circuit; reordering the filters based at least in part on the number of operations; and parallelizing the general graph such that the one or more filters are configured to be executed on one or more processors.
1. A method for accessing memory without using software-based locking techniques, the method comprising:
assigning an index number to each memory slot of a plurality of sequential memory slots of a shared non-circular memory array, wherein the index number assigned to each memory slot is based at least in part on a physical address of memory associated with that memory slot;
allocating a first memory slot and writing data to the first memory slot, wherein the first memory slot is associated with a producer index number;
reading data from a second memory slot, wherein the second memory slot is associated with a consumer index number; and
de-allocating data from a third memory slot having a second index number, wherein the third memory slot is associated with a garbage collector index number;
wherein the writing, reading and de-allocating are performed using atomic instructions.
2. The method of claim 1 , further comprising:
in accordance with a determination that a difference between the producer index number and the garbage collector index number does not exceed a maximum queue length threshold, writing data to a fourth memory slot, wherein the fourth memory slot is subsequent to the first memory slot in the sequential non-circular array.
3. The method of claim 2 , further comprising:
in accordance with a determination that a difference between the producer index number and the garbage collector index number exceeds a maximum queue length threshold, refraining from writing data to the fourth memory slot.
4. The method of claim 1 , further comprising:
in accordance with a determination that the consumer index number does not equal or exceed the producer index number, reading data from a fifth memory slot, wherein the fifth memory slot is subsequent to the second memory slot in the sequential non-circular array.
5. The method of claim 4 , further comprising:
in accordance with a determination that the consumer index number does equal the producer index number, refraining from reading data from the fifth memory slot.
6. The method of claim 1 , further comprising:
in accordance with a determination that the garbage collector index number does not meet or exceed the consumer index number, de-allocating data from a sixth memory slot, wherein the sixth memory slot is subsequent to the third memory slot in the sequential non-circular array.
7. The method of claim 6 , further comprising:
in accordance with a determination that the garbage collector index number does not meet or exceed the consumer index number, refraining from de-allocating data from the sixth memory slot.
8. The method of claim 1 , wherein the allocating includes:
retrieving a bitmap portion from a bitmap, wherein the bitmap identifies whether a state record for a queue thread state is allocated;
determining, using a single atomic instruction, whether the bitmap portion includes a bit that indicates that a memory block is available for allocation;
if the bitmap portion includes a bit that indicates that a memory block is available for allocation, performing a series of atomic instructions, per bit, to identify the bit that indicates that the memory block is available for allocation;
after identifying the bit, allocating the memory block to correspond to the first memory slot in the memory array; and
de-allocating the memory block by identifying a global position in the bitmap of the bitmap portion and subsequently identifying a bit position in the identified bitmap portion that corresponds to the previously-identified memory block.
9. The method of claim 8 , wherein the bitmap portion comprises a word.
10. A computing system including a processor and a memory, the memory storing a shared non-circular memory array and instructions which, when executed by the processor, cause the computing system to:
assign an index number to each memory slot of a plurality of sequential memory slots of a shared non-circular memory array, wherein the index number assigned to each memory slot is based at least in part on a physical address of memory associated with that memory slot;
allocate a first memory slot and writing data to the first memory slot, wherein the first memory slot is associated with a producer index number;
read data from a second memory slot, wherein the second memory slot is associated with a consumer index number; and
de-allocate data from a third memory slot having a second index number, wherein the third memory slot is associated with a garbage collector index number;
wherein the writing, reading and de-allocating are performed using atomic instructions.
11. The computing system of claim 10 , wherein the memory comprises a plurality of memory devices and the processor comprises a plurality of processors, and the shared non-circular memory array is stored on at least one of the plurality of memory devices.
12. The computing system of claim 10 , wherein the instructions include a producer process, a consumer process, and a garbage collector process.