throughput vs latency, what to look at?

[Aleksey Shipilev]

On 4/15/21 5:05 PM, Maneesh Bhunwal wrote:
> I have a generic question and a specific question. can you please help me
> with this?
> 
> JMH reports 2 gives out metrics throughput (ops/ms) and latency (ms/ops).

Note that JMH has two _benchmark modes_, not metrics: throughput and average time (not latency).

For example, the ways the results are aggregated depend on the benchmark mode. The average times are 
averaged across measurement threads, while throughput is summed up. This way, if you perform the 
ideal test with different number of threads, then average time is always C sec/op, and throughput is 
C*N ops/sec, where C is workload constant and N is the number of threads.

> When we are looking at results, which metric should we focus on? Is there
> any general guideline?

The general guideline is: it depends. This is why these two benchmark modes exist.

> I am not able to understand how to interpret the results. IMO if there are
> no locks and similar latency are there, throughput should be the same. So I
> am confused.

Is it "IMO", or have you actually profiled the workload to see there is no funky locking involved?

If you want more help, you need to show us a MCVE.

-- 
Thanks,
-Aleksey