Рет қаралды 76
Ai summarization of the weekly Nano Cryptocurrency developer update.
lyrics:
Performance Regression Investigation (V26 vs. V27)
-Issue Observed. A performance regression between V26 and V27, evident on cloud servers with lower specs, specifically when using LMDB. -Fair Queuing Mechanism. Bob suggests the regression isn't actually a regression but a difference between RocksDB and LMDB, magnified by the new fair queuing implemented in V27 to prevent message queue spamming. -Testing Results. Increasing fair queue values in V27 to match V26's behavior eliminates the throughput drop; RocksDB consistently outperforms LMDB in both throughput and latency under stress-test conditions. -Plan Forward. Given the substantial advantage of RocksDB, further efforts on optimizing LMDB are deemed less valuable. RocksDB vs. LMDB -Performance Discrepancies. Tests highlight RocksDB’s superior performance over LMDB, which wasn’t as apparent in previous versions. -RocksDB Update to Version 9. Successful initial tests of RocksDB v9 compatibility (both from scratch and upgrade paths) suggest moving to the latest version. -Documentation and Default Settings. RocksDB, still marked as experimental, is now mature enough for broader usage; discussions on making it the default database for new nodes are ongoing. General V27 Preparations -Final Patches. Remaining patches are being integrated before moving V27 to the beta testing phase. -Node Traffic and Load Management. Enhancements include reducing voting traffic by 90% through some vote broadcasting changes, and improving networking classes. -Future Optimizations and Refactoring. Future plans for refactoring of networking classes in preparation for upcoming traffic shaping implementations. Scalability and Future Prospects -Technical Debate on Network TPS. Discussion on the network’s theoretical TPS limits, with perspectives pointing to bandwidth as the primary constraint. -Parallelization Capabilities. Potential strategies for handling CPU load and IO operations include using GPUs for signature processing and employing distributed databases. -Test Network Insights. Simulations on high-core-count machines show promise, indicating IO and bandwidth are critical factors for scaling up. Miscellaneous Technical Aspects -Historical Context of Database Choices. The pause in RocksDB adoption was due to previous corruption issues; recent tests indicate stability. -Ongoing Merges and Updates. Continuous updates to upstream code bases and optimizations are ensuring smoother future integration and performance. Meeting Coordination and Closing Remarks, -Next Steps. Continued work on upstream merges, expected completion in a few weeks. -Next Meeting Logistics. Notification of potential change in hosting duties for the coming week due to availability. Nano! Nano! Your progress sings like a sweet soprano Nano! Nano! Your development plays like a grand piano