Is it possible to put the journal file to non-volatile RAM
(1) By anonymous on 2021-08-17 07:43:35 [link]
I have a hardware that supports a limited size of non-volatile RAM. So, I'd like to utilize it in Sqlite database. I found that the journal_mode can be set to MEMORY which means the journal file will be stored to the memory. Can I specify this memory address and size ? For example, the NVRAM address is mapped to the memory address 0x10000 and its size is 512KB. So, can I set the journal file to store at the address 0x10000 and the jouranl size limit be set to 512*1024 bytes ?
(2) By Simon Slavin (slavin) on 2021-08-17 12:12:26 in reply to 1
You can't do that, but there's probably a way to get the overall result you want. What is the point in setting SQLite to store a journal file in memory, but then moving that memory to non-volatile memory ? Doesn't that have the same result as just leaving the journal file in the same folder as the database, which is the default ? What result as you searching for ? And what kind of hardware are you using ? A normal desktop computer, a smartphone, or an embedded processor ?
(3) By anonymous on 2021-08-17 14:35:16 in reply to 2 [link]
>What is the point in setting SQLite to store a journal file in memory, but then moving that memory to non-volatile memory ? The purpose is to keep the behavior of memory journal mode to be the same as persist or delete mode. >Doesn't that have the same result as just leaving the journal file in the same folder as the database, which is the default ? They are a bit difference in performance. Since I'm using an embedded system, the memory journal file mode will be faster than any other mode. In embedded system, it usually perform a power cycle to reboot the hardware. In this case, I want the sqlite db file not being corrupted after a power cycle. However, the sqlite doc says that if using memory journal mode, it will a little risk for the db file safety and integrity. So, my idea is to use a NVRAM to store the memory journal file. This will provide both performance and db file safety. Is it possible ?