As far as I know memory usage between the two table types is roughly the same. The way memory is setup/used is somewhat different however. For myisam the primary memoy buffer to accelerate queries is the key_buffer which caches data for keys. In innodb you have more options to set with the main one being the innodb_buffer_pool_size which is used for caching keys and data, you want to set this as large as possible. You also have several other adjustable buffers inlcuing an 'additonal_mem_pool' which I'm not quite sure what it is used for, and the log_buffer which is used for transaction related memory I believe.
So, if you are going to be using both MyISAM and InnoDB you will need seperate buffers, which will of course increase total memory usage, or leave a smaller size for both. But if you switch completely to InnoDB you can drop the MyISAM buffers down to almost nothing (still need them as the mysql table with user data etc uses them, but say 8megs would be plenty). John On Sun, 2004-10-10 at 10:51 +0200, Jacques Jocelyn wrote: > > > Hello John, > > Interesting post, quite useful, > Question about performance with InnoDB ? > say you have a hosting server with 256 Mb of ram, would you know if > that will make a difference if the major database is converted from > MyIsam to InnoDb ? > > Although, InnoDB is not a requirement, just luxury, but I would love > to enjoy foreign keys and transactions > > Please advise, > Thanks > > > Sunday, October 10, 2004, 8:39:15 AM, you wrote: > > JM> I meant 'No transaction support', which is you can't use > JM> begin work; ... ; commit; etc to perform transactions, each query > JM> takes effect immeiately and is visible to all other > JM> threads/clients immediately. > ... > > JM> Concurrency refers to multiple seperate connections (threads) > JM> trying to read/write to/from the same table at the same time. > JM> Imagine you have 100 different connections to the database all > JM> trying to write to the same table. With MyISAM each one will lock > JM> the entire table, and only one will execute at a time, making it > JM> very slow. In InnoDB each one will only lock the rows it is > JM> modifying and they can all execute at once (if they are not > JM> modifying the same rows), and it will be very fast. > > > > > Best regards, > Jacques Jocelyn > -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
