Hi!
>>>>> "Tim" == Tim Bunce <[EMAIL PROTECTED]> writes:
Tim> On Tue, Jan 30, 2001 at 01:13:43PM -0800, Jeremy D. Zawodny wrote:
>> On Tue, Jan 30, 2001 at 05:08:46PM +0000, Tim Bunce wrote:
>> > On Tue, Jan 30, 2001 at 10:57:16PM +0800, Sam Wong wrote:
>> > > >
>> > > > So why isn't the --with-raid option set in the binary download version?
>> > >
>> > > The file will be bigger and slower in result
>> >
>> > I doubt it would be significantly bigger. I believe --with-raid
>> > support is a very small and simple layer between mysqld and the file
>> > i/o system calls.
>>
>> At the open source database summit, I asked Monty why it wasn't
>> enabled in the binaries they provide, and he said that it was for
>> performance reasons. On each table open, MySQL has to see if it is a
>> RAID table and do a bit of extra work.
Tim> Why not determine it by checking only if the open of the non-raid fail
Tim> failed? That way there'd be no penalty for those not using the raid
Tim> feature.
The problem is not just the open; The problem is that we currently have to map
all read/writes to a handler that checks if the table is raid or not.
What we need in the future is to add a wrapper around files to be able
to do things depending on things like raid and encryption on tables
without any overhead.
>> > I'm not sure what you mean by "slower in result".
>>
>> He has estimated it as a performance hit of a "few percent", but if
>> your tables stay open (because you don't have many, or you have a
>> good-sized table cache) it really shouldn't be an issue.
Tim> I can't see where a "few percent" hit would come from unless tables
Tim> were being opened at the rate of many per second. Anyone smart enough
Tim> to be using the raid feature would also know how to tune mysql to keep
Tim> the files open. Those not using the feature should see no hit.
The open is not a problem; The performance hit is for every
read/write to any data file descriptor.
Tim> Anyway, even if there is a hit, why not allow a runtime config option
Tim> to disable the raid feature? Checking a global in C is basically free.
It's still a lot of extra function calls.
>> He said that MaxSQL would have many of the compile-time options
>> enabled for folks who still wanted to use a binary release. But it
>> doesn't appear to have materialized...
Tim> I wish it would.
Tim> Monty, can you give us an update?
The problem is that our web site needs a REAL facelift as we get way
too many questions from users that can't find things they are looking
for. We will on the next 2 weeks concentrate on correcting this. After
this we will start building a mysql-max binaries which will include
support for most MySQL extensions.
Regards,
Monty
---------------------------------------------------------------------
Before posting, please check:
http://www.mysql.com/manual.php (the manual)
http://lists.mysql.com/ (the list archive)
To request this thread, e-mail <[EMAIL PROTECTED]>
To unsubscribe, e-mail <[EMAIL PROTECTED]>
Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
