But this latency already includes the maximum time a GC cycle could need to finish its work.
If without GC your application lasts 1 second and at the worst case your GC cycle lasts 0.01s, then you can be sure your entire application with GC will last no more than (1 + n*0.01) seconds, with n cycles. If this meets your real-time requirements, then it's okay. If it doesn't, you have to change your requirements, your algorithms, and so on.
On Wed, Oct 27, 2010 at 1:45 PM, Rob Kendrick
<rjek@rjek.com> wrote:
On Wed, Oct 27, 2010 at 10:37:09AM +0100, Paul Hudson wrote:
> Every hard real time system has some latency. It depends on the
> requirement if it works for your situation. If it has an upper bound
> it is hard real time in my view
The issue is that even if you can calculate the latency, there's no way
of guarenteeing you'll meet it unless you put off finishing the GC cycle
until next time.
B.
