Re: Table as fifo

[Zsolt Kendi]

RJ:"key_buffer_size: 8 -- OUCH! "
RJ: ""~100 row/sec" -- Without caching, that's about as fast as a typical disk can go. "
KZS:"The storage files located in ramdisk"

If I remember two years ago I talk about this theme (here?). The conclusion was if the index storage file located in RAMDISK and the index changed frequentally or the index size more times higher than cache then key_buffer lost their efficency. This is why practically disabled...

RJ:"please give me more details on the table, the INSERTs, and the SELECTs."
The insert is INSERT INTO [desttbl] SELECT * FROM [soutbl] WHERE [indexed datetime condition] LIMIT 1
The delete is DELETE FROM [desttbl] WHERE [indexed datetime condition] LIMIT 100
Again at this phase is no SELECT used.