A study on the Blockchain Sync Promise links accounting task scheduling with cross-ledger checks in one real-time system. Xiaoo Liu and Weiwei Bao of Heilongjiang Vocational College report that their framework reached 91.3 percent scheduling accuracy, 88.0 percent synchronization accuracy and a 1.7-second average response time in tests using 5,000 enterprise-derived accounting tasks.

The study report treats ledger consistency as feedback for the scheduler instead of a separate reconciliation step after tasks finish. The authors say that design helps the system respond to concurrent workloads while coordinating computing resources and ledger updates.

How the Blockchain Sync Promise works

The pipeline divides incoming transaction records and ledger updates into validation, aggregation and synchronization stages. A distributed message queue buffers that work, while the scheduler monitors YARN resource status, queue priority and the urgency of ledger synchronization.

Validation and aggregation tasks go to stream executor workers. Cross-ledger commits go to dedicated synchronization workers. A retry and backoff mechanism handles transient execution failures between the queue and scheduler.

A deep reinforcement learning agent assigns tasks to high, medium or low priority queues and chooses either single-container or multi-container execution on YARN. Its 24-part state combines queue statistics, CPU, memory and I/O utilization, task estimates and historical latency with indicators such as pending ledger commits.

Graph neural network and long short-term memory components add information about task dependencies and changing execution times. The graph model tracks dependencies among tasks, while the time-series model learns execution patterns from historical traces.

Permissioned ledger handles cross-ledger checks

The framework uses a permissioned consortium blockchain. Its network has one ordering node, four peer nodes and two ledger gateway nodes connected to Spark Streaming workers. The authors chose Raft-based consensus to avoid mining delays.

When a ledger update reaches a synchronization worker, a smart contract registers transaction metadata, ledger identifiers, version numbers, hash digests, priority labels and timeout thresholds. If commit conditions are met, a zero-knowledge proof module creates a non-interactive consistency proof over hash commitments and range constraints for sensitive accounting fields. Validators check the proof and public commitments without seeing the underlying values.

The test dataset came from eight medium- and large-scale enterprises that piloted or deployed the system between January and December 2023. It contained 5,000 anonymized task instances: transaction validation made up 44 percent, balance aggregation and reconciliation 30 percent, report generation 16 percent, and cross-ledger synchronization 10 percent.

Researchers created high-concurrency scenarios by compressing arrival intervals for those records. They describe this as a replay protocol that increases queue pressure and ledger contention without adding new labeled samples.

Benchmark results and stated limits

A static scheduler recorded 78.2 percent scheduling accuracy, 74.5 percent synchronization accuracy and a 2.1-second response time. A conventional deep reinforcement learning scheduler without graph modeling or blockchain feedback reached 85.6 percent, 81.3 percent and 1.9 seconds. The full system achieved the best results across those measures.

Removing reinforcement learning, graph dependency modeling, or blockchain verification worsened response time by 0.28, 0.17, and 0.09 seconds, respectively. A version without live ledger feedback also performed worse.

Average CPU utilization was 74.6 percent and peak memory utilization was 68.9 percent. Network traffic rose from 12.3 to 18.7 megabytes per second, with cryptographic checks limited to the 10 percent of cross-ledger tasks.

Fault tests included peer interruptions, network degradation, ledger contention and workload drift. Recovery took 12.6 to 18.5 seconds; under ledger contention, synchronization accuracy fell 7.6 percentage points and rollback rose to 8.7 percent.

The authors caution that the dataset is limited and replayed concurrency adds no independent observations. Results describe the tested cluster and protocol; validation across clusters, platforms and regions remains future work, as does adding network-state features.

For related work on blockchain verification and enterprise ledger systems, Coin Review News has covered DIFC court record verification, tokenization settlement infrastructure and cross-chain trading verification.

The authors identify broader validation as the next step.