Systems and methods for implementing persistent data structures on an asymmetric non-volatile memory architecture

A computing system, comprising:

• a front-end node comprising:
• at least one central processing unit (CPU), a dynamic random-access memory (DRAM) and a first Remote Direct Memory Access (RDMA) enabled network interface card (RNIC);
• and
• a back-end node comprising a second RNIC and a non-volatile memory (NVM),
• wherein the at least one CPU at the front-end node is configured to:
• generate a data structure operation record for a data structure operation, the data structure operation being directed to a data structure persisted in the NVM and at least partially cached in the DRAM, the data structure operation record including an operation type and parameter(s) for the data structure operation;
• append the data structure operation record in an operation log;
• generate a transaction record for a transaction, the transaction including a plurality of memory operations that collectively accomplishing the data structure operation on the data structure persisted in the NVM, the transaction record including a plurality of memory operation records corresponding to the plurality of memory operations;
• append the transaction record in a transaction log;
• flush the operation log to the back-end node via the first RNIC;
• and
• flush the transaction log to the back-end node via the first RNIC after flushing the operation log;
• wherein the second RNIC at the back-end node is configured to:
• persist the operation log received from the front-end node in the NVM;
• persist the transaction log received from the front-end node in the NVM;
• and
• perform memory updates according to the plurality of the memory operation records to accomplish the data structure operation, with the data structure operation record in the operation log received by the back-end node serving as a commit signal for the transaction.

One or more computer-readable non-transitory media comprising one or more instructions that when executed by one or more processors is to configure the one or more processors to cause performance of operations comprising:

• at a front-end node in a computing system:
• generating a data structure operation record for a data structure operation, the data structure operation being directed to a data structure persisted in a non-volatile memory (NVM) in a back-end node of the computing system, the data structure operation record including an operation type and parameter(s) for the data structure operation;
• appending the data structure operation record in an operation log;
• generating a transaction record for a transaction, the transaction including a plurality of memory operations that collectively accomplishing the data structure operation on the data structure persisted in the NVM, the transaction record including a plurality of memory operation records corresponding to the plurality of memory operations;
• appending the transaction record in a transaction log;
• flushing the operation log to the back-end node;
• and
• flushing the transaction log to the back-end node after flushing the operation log;
• and
• at the back-end node of the computing system:
• persisting the operation log received from the front-end node in the NVM;
• persisting the transaction log received from the front-end node in the NVM;
• and
• performing memory updates according to the plurality of the memory operation records to accomplish the data structure operation, with the data structure operation record in the operation log received by the back-end node serving as a commit signal for the transaction.