Java: low latency performance measurements

Question

Is the process of measuring performance for low latency Java apps different from that of measuring performance for non low latency Java apps? If yes, how so, and are there any specific tools used for measuring low latency performance?

EDIT: e.g. here's a concrete piece of information: http://www.artima.com/forums/flat.jsp?forum=121&thread=212810

Answer 1

The main difference with low latency timings is that

• every micro-second counts. You will have an idea of most much each micro-second costs your business per year and how much time it is worth reducing each micro-second.
• you want to measure the highest 99% or even 99.99% latencies. (worst 1% or 0.01% respectly)
• you want a fast clock which is often limited to one host, or even one socket. (You can measure low latency between hosts with specialist hardware) For multi-millisecond timings you can relatively easily measure between hosts (with just NTP configured)
• you want to minimise garbage, esp in your measurements.
• it is quite likely you will need to develop application specific tools which are embedded into the application and run in production. You can use profilers as a start but most ultra low latency applications don't show anything useful in commercial profilers (nor do they GC much, if at all when running)

You can have a read of my blog for general low latency, high performance testing practices (some of these are nano-second based). Vanilla Java