* second" (commonly referred to as <i>QPS</i>, queries per second).
   *
   * <p>The returned {@code RateLimiter} ensures that on average no more than {@code
   * permitsPerSecond} are issued during any given second, with sustained requests being smoothly
   * spread over each second. When the incoming request rate exceeds {@code permitsPerSecond} the

    assertTrue(limiter.tryAcquire(0, SECONDS));
    assertFalse(limiter.tryAcquire(0, SECONDS));
    assertFalse(limiter.tryAcquire(0, SECONDS));
    stopwatch.sleepMillis(100);
    assertFalse(limiter.tryAcquire(0, SECONDS));
  }

  public void testTryAcquire_someWaitAllowed() {
    RateLimiter limiter = RateLimiter.create(5.0, stopwatch);
    assertTrue(limiter.tryAcquire(0, SECONDS));

    assertTrue(limiter.tryAcquire(0, SECONDS));
    assertFalse(limiter.tryAcquire(0, SECONDS));
    assertFalse(limiter.tryAcquire(0, SECONDS));
    stopwatch.sleepMillis(100);
    assertFalse(limiter.tryAcquire(0, SECONDS));
  }

  public void testTryAcquire_someWaitAllowed() {
    RateLimiter limiter = RateLimiter.create(5.0, stopwatch);
    assertTrue(limiter.tryAcquire(0, SECONDS));

   *
   * How this works is best explained with an example:
   *
   * For a RateLimiter that produces 1 token per second, every second that goes by with the
   * RateLimiter being unused, we increase storedPermits by 1. Say we leave the RateLimiter unused
   * for 10 seconds (i.e., we expected a request at time X, but we are at time X + 10 seconds before
   * a request actually arrives; this is also related to the point made in the last paragraph), thus

    // stop it.
    blockingCallable.waitForStart();

    // Give the second task a chance to (incorrectly) start up while the first task is running.
    assertThat(future2.isDone()).isFalse();

    // Stop the first task. The second task should then run.
    blockingCallable.stop();
    executor.shutdown();
    assertThat(executor.awaitTermination(10, SECONDS)).isTrue();
    assertThat(getDone(future2)).isFalse();
  }

        });

        // First resolve
        deferred.resolve(new SuggestResponse("first", 0, Collections.emptyList(), 0, null));
        // Second resolve (should be ignored)
        deferred.resolve(new SuggestResponse("second", 0, Collections.emptyList(), 0, null));
        // Third resolve (should be ignored)
        deferred.resolve(new SuggestResponse("third", 0, Collections.emptyList(), 0, null));

    // stop it.
    blockingCallable.waitForStart();

    // Give the second task a chance to (incorrectly) start up while the first task is running.
    assertThat(future2.isDone()).isFalse();

    // Stop the first task. The second task should then run.
    blockingCallable.stop();
    executor.shutdown();
    assertThat(executor.awaitTermination(10, SECONDS)).isTrue();
    assertThat(getDone(future2)).isFalse();
  }

					b.ReportMetric(float64(n)*float64(ops)/spent.Seconds(), "vops/s")
					b.ReportMetric(latency/float64(ops), "ms/op")
				}
			})
		}
	})
	b.Run("rpc", func(b *testing.B) {
		for par := 1; par <= 32; par *= 2 {
			b.Run("par="+strconv.Itoa(par*runtime.GOMAXPROCS(0)), func(b *testing.B) {
				defer timeout(60 * time.Second)()
				ctx, cancel := context.WithTimeout(b.Context(), 30*time.Second)
				defer cancel()
				b.ReportAllocs()

      return;
    }
    long timeoutSeconds = timeoutSeconds();
    long deadline = System.nanoTime() + SECONDS.toNanos(timeoutSeconds);
    do {
      System.runFinalization();
      if (future.isDone()) {
        return;
      }
      System.gc();
      try {
        future.get(1L, SECONDS);
        return;
      } catch (CancellationException | ExecutionException ok) {
        return;

      return;
    }
    long timeoutSeconds = timeoutSeconds();
    long deadline = System.nanoTime() + SECONDS.toNanos(timeoutSeconds);
    do {
      System.runFinalization();
      if (future.isDone()) {
        return;
      }
      System.gc();
      try {
        future.get(1L, SECONDS);
        return;
      } catch (CancellationException | ExecutionException ok) {
        return;