unix multi-user DBMS, questions re. data transfer

[3131]

hi my name is jeff martin,

i am developing a multi-user DBMS for Unix operating systems.
in doing so i have learned quite a bit about IPCs and their
ability to communicate and transfer data between processes.
our DBMS takes advantage of multi-process logic in retrieving
data from the database and distributing to users as requested.

currently i am using a combination of semaphores and shared
memory to accomplish the data transfer. however i am unsatisfied
with the complexity of the communication logic. i am also
curious about my options for increased transfer speed.

what other options are available? how do these options compare
in speed and complexity? pipes versus shared memory? signals versus
semaphores? we have avoided signals because of the 'sdb' problem
in dealing with signals. does a debugger exist that can handle
signals? 

i desire to increase my data transfer speed and decrease my
code's complexity. i'd like a few other miracles as well but lets
start with this one first.

thanks, jeff martin, programmer, phillies.