List:General Discussion« Previous MessageNext Message »
From:Heikki Tuuri Date:January 17 2002 2:37pm
Subject:Re: Memory
View as plain text  
John,

I suggest setting record_buffer to 1 MB. Disk reads in blocks of 1 MB are
probably as fast as in blocks of 10 MB.

Also set sort_buffer to 1 MB, and only increase it if there are performance
problems.

The maximum process space of Linux x86 is 2 GB, and better play safe.

Jeremy, I think some Intel x86 processors support segmented memory above > 4
GB. Is that supported in Linux?

Jeremy, also thanks for your article in the latest Linux Magazine. I too
learned something about tuning MySQL :).

Best regards,

Heikki Tuuri
Innobase Oy
---
Order technical MySQL/InnoDB support at https://order.mysql.com/
See http://www.innodb.com for the online manual and latest news on InnoDB

Jeremy Zawodny wrote in message ...
>On Wed, Jan 16, 2002 at 09:46:48PM -0500, John Kemp wrote:
>>  From the innodb.com site (bugs & fixes):
>>
>> This has me worried, but I haven't seen this behaviour on our site. We
have
>>
>> innodb_buffer_pool = 1100Mb
>> key_buffer = 400 Mb
>> record_buffer = 10Mb
>> sort_buffer = 20Mb
>> max_connections = 220
>>
>> which according to this formula gives me 1100 + 400 + (220 * (20 + 10))
>> + ( 220 * 1)  = 8320Mb at max capacity. At roughly half capacity (96
>> connections) we're using only 1390Mb, so I'm finding it hard to believe
>> it's going to scale that badly right now. We "only" have 4Gb memory on
>> our linux-based database machine right now - should I be upgrading? ;-)
>>
>> Does anyone have any information that either supports or refutes the
>> statement above? I'd be interested if you did....
>
>Well, the sort_buffer and record_buffer will only be allocated on an
>as-needed basis.  And they'll exist for very short periods of time,
>ideally.
>
>So your 1390 comes mainly from innodb_buffer_pool + key_buffer which
>are the two "global" buffers (non-thread-specific) that are involved.
>You'll likely see a single MySQL thread peek above that on occasion,
>but you'd need things to get pretty bad before you eat up all your
>memory.
>
>See
>
>  http://jeremy.zawodny.com/mysql/mysql2.pdf
>
>for a bit of info on the difference between global and per-thread
>memory in MySQL.
>
>Jeremy
>--
>Jeremy D. Zawodny, <jzawodn@stripped>
>Technical Yahoo - Yahoo Finance
>Desk: (408) 349-7878   Fax: (408) 349-5454   Cell: (408) 685-5936
>
>MySQL 3.23.41-max: up 14 days, processed 336,064,611 queries (270/sec. avg)


Thread
MemoryJohn Kemp17 Jan
  • Re: MemoryJeremy Zawodny17 Jan
Re: MemoryHeikki Tuuri17 Jan
  • Re: MemoryJohn Kemp17 Jan
  • Re: MemoryJeremy Zawodny17 Jan
    • Re: MemoryDan Nelson18 Jan
      • Re: MemoryJeremy Zawodny18 Jan
Re: Memoryalec.cawley17 Jan
Re: MemoryHeikki Tuuri17 Jan