Using mmap you share memory among processes. You have at least two ways: 1) named 2) unnamed
With 1) named mmap you just open the handle in each process. With 2) you need to open the handle before the fork. I suggest 1) because fcgi or scgi fork early so you could not be able to create an unnamed mmap before fork, even if it is possible. Locking/semaphores is not a problem, just think of using the same semantic of file locking (i.e. each process can lock for read or write random ranges of bytes). You can reach a very high level of parallelism with lower dead-locking probability than with java threads. mic mic 2012/2/8 Ross Peoples <ross.peop...@gmail.com>: > That is a VERY good question. If you already have a setup like this, give it > a shot and see if it works....I would be interested to know the result.
