Research & Writing

By Jumail J · Systems Architecture

High-performance databases often rely on memory-mapped files (`mmap`) to delegate file I/O caching to the kernel. However, garbage-collected runtimes like Go introduce latency spikes when coordinating memory-mapped virtual allocations with runtime heap compaction. In this paper, we analyze the structural page fault latency and design an optimal kernel-bypass memory allocator.

By Jumail J · Systems Architecture

Distributed databases utilize materialized views to pre-compute expensive aggregations and join queries. However, maintaining view consistency under high-frequency writes incurs substantial synchronization latency. In this paper, we present an asynchronous, transactionally consistent refresh engine powered by WAL-based log parsing and lock-free event pipelines.

By Jumail J · Theory & Logic

Lock-free graph traversal algorithms are highly performant but prone to concurrency anomalies like cycles, race conditions, and dangling pointers. In this work, we present a formal verification methodology for concurrent, lock-free Breadth-First Search (BFS) using temporal logic specifications and Automated Theorem Proving.