Re: memory mapped store entries

Anthony Baxter wrote:

>>>> "Stephen R. van den Berg" wrote
>> >Hmm. Isn't there one swaplog per cache disk?

>> Yes, but you can configure where they should be put. I put all
>> swapfiles together on a separate disk.

>Um, under load won't that mean that one disk gets somewhat overloaded?

Depends on the load. In the worst case, there is going to be one disk
write per new store entry.
But since the updates are being cached, and coalesced when within the
caching timeframe several updates hit the same page, there'll be a lot less.
The nice thing is that the kernel is allowed to defer the updates
indefinitely. Since the memory is present, it normally never is forced
to update. It will depend on the kernel's autosync policy regarding
mmapped files, how often it really will flush.

The ideal situation would be if we could tell the kernel never to flush
the dirty pages automatically, but force a flush using msync(MS_ASYNC)
once every night at say 5AM.

-- 
Sincerely,                                                          srb@cuci.nl
           Stephen R. van den Berg (AKA BuGless).
Sign at a nursery school: Please pick up your children promptly after service.
                          No refunds after Wednesday.