Forum Discussion
The config exists in multiple places.
mcpd has a copy of the config in memory - this is what is used by tmm and other processes on the BigIP to process traffic. When you make a change to the config, it is made in the in-memory copy, and then written to the on-disk config files. This process does not work the other way way round - the on-disk config files are only loaded into mcpd to become the running config if you do a "tmsh load sys config" or run some other config-reloading command.
So we have two copies of the config - in-memory or the config files. The third copy is a binary config written to disk by mcpd. This is a direct dump of the in-memory data structure on to disk. This is much faster to load than parsing the config files, so when mcpd is starting, it uses this (if it exists). It's the difference between a few seconds to load the config, or minutes to parse and load the config files.
So if you change bigip.conf or bigip_base.conf, you will need to reload the config to see the change have effect. If you make a change in the WebUI, mcpd makes the change in running memory, and then writes the running config back into the config files on disk (both the binary and text config files) overwriting any pending changes that may have been in those files.