Your task is to improve a multi-threaded program that frequently writes small entries to a shared file by introducing a layer that merges in memory before writing to file to reduce the number of expensive I/O operations. Imagine a program which writes to file small log entries very frequently. The program is multithreaded and each thread writes log entries to a common log file. To simplify implementation, each thread just writes log individually with blocking I/O interface. Even though threads write to the shared file concurrently without any locking, this is correct because the OS guarantees the atomicity of individual write() operation