Inconsistent Execution Time of RAND_bytes in OpenSSL during Benchmarking

I'm benchmarking the RAND_bytes function from the OpenSSL library in C for cryptographic operations. However, I've encountered an issue where the execution time for this function is highly inconsistent. Every time I run the benchmark, the elapsed time varies significantly. This variability makes it challenging to obtain a consistent benchmark result. Is this variability in execution time an expected behavior for RAND_bytes?

Here's a simplified version of my code that measures the time taken:

#include <stdio.h>
#include <openssl/rand.h>
#include <sys/time.h>

// Function to get the current time in microseconds
long long current_time_microseconds() {
    struct timeval tv;
    gettimeofday(&tv, NULL);
    return (long long)tv.tv_sec * 1000000LL + tv.tv_usec;
}

int main() {
    unsigned char buffer[128];
    long long start_time, end_time;

    start_time = current_time_microseconds();
    if (RAND_bytes(buffer, sizeof(buffer)) != 1) {
        // Handle error
        fprintf(stderr, "Error generating random bytes.\n");
        return 1;
    }
    end_time = current_time_microseconds();

    printf("Time taken (microseconds): %lld\n", end_time - start_time);
    return 0;
}

Is there an inherent reason within the RAND_bytes function or its implementation in OpenSSL that might cause such variability in execution time? And is there any way to mitigate this inconsistency to get more stable benchmark results?