4

I have a linux machine, on which a ruby web server and a c++ process are hosted. Both the components work on the same database.

In order to avoid repeated database requests, I want to make and maintain an object model.

Is it possible to make one object model and share it between ruby server and c++ process. If yes what can be the pros and cons?

Schu
  • 1,124
  • 3
  • 11
  • 23
  • 1
    Take a look at this question http://stackoverflow.com/questions/1542520/how-can-i-call-c-functions-from-within-ruby – Oscar Del Ben Jul 02 '12 at 15:08
  • calling c++ api from ruby is an option. I was initially concerned by the number of api functions I have to make, but may be its better that way as I can control whats exposed. – Schu Jul 02 '12 at 15:43

2 Answers2

3

In short; you cannot read the memory of a separate arbitrary process. If you are able to recompile both programs, you can implement memory sharing/IPC using ptrace.

I would, however recommend you look into memcached. It runs as a daemon (locally if you like) and provides a very fast name-value cache that you can use for anything. We use it for caching the results of expensive calculations on a rails system.

lynks
  • 5,599
  • 6
  • 23
  • 42
  • 2
    I would recommend giving Redis a try also. It's a key-value in-memory storage similar to Memcached. – Samy Dindane Jul 02 '12 at 15:34
  • memcached looks good. But in my application, I need something that can take the place of the database. – Schu Jul 02 '12 at 15:37
  • You can greatly reduce the number of database hits by storing the results in Memcached (against something as simple as a hash of the SQL). But if you are wanting to get rid of the database entirely, Memcached (as a cache) is not the tool for the job at all. You're unlikely to find something that does 'being a database' better than a database :P – lynks Jul 02 '12 at 15:43
2

If you really want to share memory between processes, then look for the shared memory (man shm.h) and semaphores (man sem.h) in the POSIX Programmer's Manual. It allows you to share a raw memory between C++ and Ruby. It is not the simplest way - you need to write FFI to shm.h and sem.h in Ruby and implement appropriate protocol over the shared memory buffer (using semaphores for synchronizing).

You can also use some other IPC facilities like pipes, sockets, RPC's, MQ's and even DB's, but then it is copying rather than sharing.

JJJ
  • 2,731
  • 1
  • 12
  • 23