Dear list members, Recently i have encountered a problem with system resources on my Linux (64 bit) machine with Oracle 11g. I have started 2 instances on the same server with 1GB of MEMORY_TARGET each. We have run some sort of stress test trying to open 200-300 application connections to the database and suddenly got many errors "too many open files". As you know Oracle 11g uses a different mechanism for memory management implementing shared memory on so called tmpfs file system or /dev/shm/. In previous versions shared segment can be easily observed from ipcs -m command but in Oracle 11g the situation is different , many small files of 4MB (MEMORY_TARGET under 1GB, and 16MB for MEMORY_TARGET above 1G) are created under /dev/shm. So i tried to understand how all this is related to "too many open files error". I have tried to examine lsof output and found that ~55000 files are opened. After that i did a small experiment: 1. shutdown the database ; /dev/shm is empty 2. startup the database : found 250 files (makes sense 1GB of memory_traget / 4M =250 files) lsof gives me output 5000 open files , without any opened external sessions ,which means that all files were opened by Oracle background processes like pmon, dbwr, lgwr ... I have checked this one more time and yes indeed i found that by default my instance is started with 20 background processes. The most interesting point that each Oracle process opens every file under /dev/shm that has been created by Oracle, 20 processes * 250 files =5000 - this is how i got 5000 open files without even opening a single connection to Oracle. 3. After that i run a simple shell script that opens 200 sessions without closing them. The number of open files immediately jumped to 55000. From my first tests i can see that number of opened files by Oracle is directly related to MEMORY_TARGET size of each instance * number of Oracle processes. Taking into account that every OS has it's own limits for maximum number of possible open files new memory management approach looks to me a little bit problematic. I would like to know if someone has experiences the same and how you tried to solve this problem. Is there any option to increase the basic file size under /dev/shm for example ? Thank you. -- Best Regards Michael Elkin