Using Virtual Memory for Process Configuration Data
Configuration data for each process in Junos OS is stored
in memory that is mapped within the address space of each process,
requiring a fixed maximum space to be reserved in each process. This
scheme works well until a process is managing many functions at commit
time and negatively impacts the commit time, or simply needs more
memory than the default allotment. For example, the rpd
process might be managing many routes and require more space to
store important information about the routes.
In circumstances that require more than the maximum memory-mapped
size, you can use virtual-memory-mapping
at the [edit
system configuration-database]
hierarchy level to make more
memory available for the configuration database per process.
You can configure a portion of virtual memory at a fixed size for the initial portion of the configuration database, and you can specify an amount to be used for page-pooling. Page-pooling uses a small amount of memory to bring database pages into memory as needed, rather than mapping the entire configuration database into the virtual memory space for the process.