如何模拟慢速更新sql？

Question

DBA告诉我，我应该优化一些慢的SQL，

实际上有10个相同的更新sql，最长的时间是56秒。

这个sql的业务逻辑是，如果用户赢得奖励，应减少该奖励的一个股票。

我认为这是由高并发引起的，只有在这种情况下才存在锁定竞争。然后我决定在我的本地环境中重新生成它，我让很多线程同时执行更新操作，但我无法重现上面的慢sql，下面是我的实验结果：

thred count|100|500|1000|2000
-----|-----|-----|-----|-----
took time(ms)|149|490|1064|1996

即使2000更新sql并发执行，它只需要1996 ms来执行所有。那么如何重新制作这些慢速sql？

我的代码就像这样

@Test
public void test_concurrent_update() throws InterruptedException {
    int nThreads = 50; //thread count

    ExecutorService pool = Executors.newFixedThreadPool(nThreads);
    final CountDownLatch startLatch = new CountDownLatch(nThreads);
    final CountDownLatch endLatch = new CountDownLatch(nThreads);
    long startTime = System.currentTimeMillis();
    for (int i = 0; i < nThreads; i++) {
        pool.submit(new Runnable() {
            @Override
            public void run() {
                startLatch.countDown(); // let all threads concurrently execute update operation
                try { startLatch.await(); } catch (InterruptedException e) { e.printStackTrace(); }
                mapper.updateAwardStock(); // reduce stock by mybatis
                endLatch.countDown();
                System.out.printf("%s down%n",Thread.currentThread().getName());
            }
        });
    }

    endLatch.await(); // wait for all threads are finished executing
    long endTime = System.currentTimeMillis();
    System.out.printf("%d threads concurrent update stock took time: %d(ms)%n",nThreads,(endTime-startTime));

}

@Update("update award_stock set stock = stock-1 where award_id = 1 and  stock>0 limit 1")
void updateAwardStock();